Merge branch 'Comfy-Org:master' into master

.ci/windows_amd_base_files/README_VERY_IMPORTANT.txt

@@ -1,5 +1,4 @@
-As of the time of writing this you need this driver for best results:
-https://www.amd.com/en/resources/support-articles/release-notes/RN-AMDGPU-WINDOWS-PYTORCH-7-1-1.html
+As of the time of writing this you need a recent driver. Updating to the latest driver is recommended.
 
 HOW TO RUN:
 
@@ -7,9 +6,9 @@ If you have a AMD gpu:
 
 run_amd_gpu.bat
 
-If you have memory issues you can try disabling the smart memory management by running comfyui with:
+If you have memory issues you can try enabling the new dynamic memory management by running comfyui with:
 
-run_amd_gpu_disable_smart_memory.bat
+run_amd_gpu_enable_dynamic_vram.bat
 
 IF YOU GET A RED ERROR IN THE UI MAKE SURE YOU HAVE A MODEL/CHECKPOINT IN: ComfyUI\models\checkpoints
 

.github/workflows/check-line-endings.yml

@@ -17,7 +17,7 @@ jobs:
       - name: Check for Windows line endings (CRLF)
         run: |
           # Get the list of changed files in the PR
-          CHANGED_FILES=$(git diff --name-only ${{ github.event.pull_request.base.sha }}..${{ github.event.pull_request.head.sha }})
+          CHANGED_FILES=$(git diff --name-only ${{ github.event.pull_request.base.sha }}..${{ github.event.pull_request.head.sha }} -- ':!.ci')
 
           # Flag to track if CRLF is found
           CRLF_FOUND=false

README.md

@@ -140,7 +140,7 @@ ComfyUI follows a weekly release cycle targeting Monday but this regularly chang
    - Commits outside of the stable release tags may be very unstable and break many custom nodes.
    - Serves as the foundation for the desktop release
 
-2. **[ComfyUI Desktop](https://github.com/Comfy-Org/desktop)**
+2. **[Comfy Desktop](https://github.com/Comfy-Org/Comfy-Desktop)**
    - Builds a new release using the latest stable core version
 
 3. **[ComfyUI Frontend](https://github.com/Comfy-Org/ComfyUI_frontend)**
@@ -309,7 +309,7 @@ After this you should have everything installed and can proceed to running Comfy
 
 #### Apple Mac silicon
 
-You can install ComfyUI in Apple Mac silicon (M1 or M2) with any recent macOS version.
+You can install ComfyUI in Apple Mac silicon (M1, M2, M3 or M4) with any recent macOS version.
 
 1. Install pytorch nightly. For instructions, read the [Accelerated PyTorch training on Mac](https://developer.apple.com/metal/pytorch/) Apple Developer guide (make sure to install the latest pytorch nightly).
 1. Follow the [ComfyUI manual installation](#manual-install-windows-linux) instructions for Windows and Linux.
@@ -364,7 +364,7 @@ For models compatible with Iluvatar Extension for PyTorch. Here's a step-by-step
 | Flag | Description |
 |------|-------------|
 | `--enable-manager` | Enable ComfyUI-Manager |
-| `--enable-manager-legacy-ui` | Use the legacy manager UI instead of the new UI (requires `--enable-manager`) |
+| `--enable-manager-legacy-ui` | Use the legacy manager UI instead of the new UI (implies `--enable-manager`) |
 | `--disable-manager-ui` | Disable the manager UI and endpoints while keeping background features like security checks and scheduled installation completion (requires `--enable-manager`) |
 
 
@@ -382,11 +382,7 @@ For AMD 7600 and maybe other RDNA3 cards: ```HSA_OVERRIDE_GFX_VERSION=11.0.0 pyt
 
 ### AMD ROCm Tips
 
-You can enable experimental memory efficient attention on recent pytorch in ComfyUI on some AMD GPUs using this command, it should already be enabled by default on RDNA3. If this improves speed for you on latest pytorch on your GPU please report it so that I can enable it by default.
-
-```TORCH_ROCM_AOTRITON_ENABLE_EXPERIMENTAL=1 python main.py --use-pytorch-cross-attention```
-
-You can also try setting this env variable `PYTORCH_TUNABLEOP_ENABLED=1` which might speed things up at the cost of a very slow initial run.
+You can try setting this env variable `PYTORCH_TUNABLEOP_ENABLED=1` which might speed things up at the cost of a very slow initial run.
 
 # Notes
 
@@ -462,16 +458,6 @@ To use the most up-to-date frontend version:
 
 This approach allows you to easily switch between the stable fortnightly release and the cutting-edge daily updates, or even specific versions for testing purposes.
 
-### Accessing the Legacy Frontend
-
-If you need to use the legacy frontend for any reason, you can access it using the following command line argument:
-
-```
---front-end-version Comfy-Org/ComfyUI_legacy_frontend@latest
-```
-
-This will use a snapshot of the legacy frontend preserved in the [ComfyUI Legacy Frontend repository](https://github.com/Comfy-Org/ComfyUI_legacy_frontend).
-
 # QA
 
 ### Which GPU should I buy for this?

alembic_db/versions/0004_drop_tag_type.py

@@ -0,0 +1,39 @@
+"""
+Drop the vestigial tags.tag_type column.
+
+tag_type was always "user" in practice — no code path ever set it to anything
+else (no system/seeded classification was ever wired up) and nothing queried it.
+The column, its index (ix_tags_tag_type), and the corresponding API field were
+dead weight, so they are removed.
+
+Revision ID: 0004_drop_tag_type
+Revises: 0003_add_metadata_job_id
+Create Date: 2026-06-03
+"""
+
+from alembic import op
+import sqlalchemy as sa
+
+revision = "0004_drop_tag_type"
+down_revision = "0003_add_metadata_job_id"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    with op.batch_alter_table("tags") as batch_op:
+        batch_op.drop_index("ix_tags_tag_type")
+        batch_op.drop_column("tag_type")
+
+
+def downgrade() -> None:
+    with op.batch_alter_table("tags") as batch_op:
+        batch_op.add_column(
+            sa.Column(
+                "tag_type",
+                sa.String(length=32),
+                nullable=False,
+                server_default="user",
+            )
+        )
+        batch_op.create_index("ix_tags_tag_type", ["tag_type"])

app/assets/api/routes.py

@@ -39,6 +39,7 @@ from app.assets.services import (
     update_asset_metadata,
     upload_from_temp_path,
 )
+from app.assets.services.cursor import InvalidCursorError
 from app.assets.services.tagging import list_tag_histogram
 
 ROUTES = web.RouteTableDef()
@@ -174,7 +175,7 @@ def _build_asset_response(result: schemas.AssetDetailResult | schemas.UploadResu
         user_metadata=result.ref.user_metadata or {},
         metadata=result.ref.system_metadata,
         job_id=result.ref.job_id,
-        prompt_id=result.ref.job_id,  # deprecated: mirrors job_id for cloud compat
+        prompt_id=result.ref.job_id,  # deprecated alias of job_id, kept for compatibility
         created_at=result.ref.created_at,
         updated_at=result.ref.updated_at,
         last_access_time=result.ref.last_access_time,
@@ -211,24 +212,37 @@ async def list_assets_route(request: web.Request) -> web.Response:
     order_candidate = (q.order or "desc").lower()
     order = order_candidate if order_candidate in {"asc", "desc"} else "desc"
 
-    result = list_assets_page(
-        owner_id=USER_MANAGER.get_request_user_id(request),
-        include_tags=q.include_tags,
-        exclude_tags=q.exclude_tags,
-        name_contains=q.name_contains,
-        metadata_filter=q.metadata_filter,
-        limit=q.limit,
-        offset=q.offset,
-        sort=sort,
-        order=order,
-    )
+    try:
+        result = list_assets_page(
+            owner_id=USER_MANAGER.get_request_user_id(request),
+            include_tags=q.include_tags,
+            exclude_tags=q.exclude_tags,
+            name_contains=q.name_contains,
+            metadata_filter=q.metadata_filter,
+            limit=q.limit,
+            offset=q.offset,
+            sort=sort,
+            order=order,
+            after=q.after,
+        )
+    except InvalidCursorError as e:
+        return _build_error_response(400, "INVALID_CURSOR", str(e))
 
     summaries = [_build_asset_response(item) for item in result.items]
 
+    # has_more semantics differ by mode:
+    #   - cursor mode: a non-empty next_cursor means there are more results.
+    #   - offset mode: derived from total - (offset + page size).
+    if q.after is not None:
+        has_more = result.next_cursor is not None
+    else:
+        has_more = (q.offset + len(summaries)) < result.total
+
     payload = schemas_out.AssetsList(
         assets=summaries,
         total=result.total,
-        has_more=(q.offset + len(summaries)) < result.total,
+        has_more=has_more,
+        next_cursor=result.next_cursor,
     )
     return web.json_response(payload.model_dump(mode="json", exclude_none=True))
 
@@ -519,18 +533,14 @@ async def update_asset_route(request: web.Request) -> web.Response:
 @_require_assets_feature_enabled
 async def delete_asset_route(request: web.Request) -> web.Response:
     reference_id = str(uuid.UUID(request.match_info["id"]))
-    delete_content_param = request.query.get("delete_content")
-    delete_content = (
-        False
-        if delete_content_param is None
-        else delete_content_param.lower() not in {"0", "false", "no"}
-    )
 
     try:
+        # Deleting an asset is a soft delete of the reference; the underlying
+        # content is preserved (it may be shared with other references).
         deleted = delete_asset_reference(
             reference_id=reference_id,
             owner_id=USER_MANAGER.get_request_user_id(request),
-            delete_content_if_orphan=delete_content,
+            delete_content_if_orphan=False,
         )
     except Exception:
         logging.exception(
@@ -575,8 +585,8 @@ async def get_tags(request: web.Request) -> web.Response:
     )
 
     tags = [
-        schemas_out.TagUsage(name=name, count=count, type=tag_type)
-        for (name, tag_type, count) in rows
+        schemas_out.TagUsage(name=name, count=count)
+        for (name, count) in rows
     ]
     payload = schemas_out.TagsList(
         tags=tags, total=total, has_more=(query.offset + len(tags)) < total

app/assets/api/schemas_in.py

@@ -59,6 +59,11 @@ class ListAssetsQuery(BaseModel):
 
     limit: conint(ge=1, le=500) = 20
     offset: conint(ge=0) = 0
+    # Opaque keyset cursor. When supplied, `offset` is ignored. Cursor pagination
+    # is supported for sort values `created_at`, `updated_at`, `name`, `size`.
+    # Supplying `after` together with `sort=last_access_time` returns
+    # 400 INVALID_CURSOR; that sort only supports offset/limit.
+    after: str | None = None
 
     sort: Literal["name", "created_at", "updated_at", "size", "last_access_time"] = (
         "created_at"

app/assets/api/schemas_out.py

@@ -41,12 +41,13 @@ class AssetsList(BaseModel):
     assets: list[Asset]
     total: int
     has_more: bool
+    # Opaque cursor for the next page. Omitted when there are no more results.
+    next_cursor: str | None = None
 
 
 class TagUsage(BaseModel):
     name: str
     count: int
-    type: str
 
 
 class TagsList(BaseModel):

app/assets/database/models.py

@@ -227,7 +227,6 @@ class Tag(Base):
     __tablename__ = "tags"
 
     name: Mapped[str] = mapped_column(String(512), primary_key=True)
-    tag_type: Mapped[str] = mapped_column(String(32), nullable=False, default="user")
 
     asset_reference_links: Mapped[list[AssetReferenceTag]] = relationship(
         back_populates="tag",
@@ -240,7 +239,5 @@ class Tag(Base):
         overlaps="asset_reference_links,tag_links,tags,asset_reference",
     )
 
-    __table_args__ = (Index("ix_tags_tag_type", "tag_type"),)
-
     def __repr__(self) -> str:
         return f"<Tag {self.name}>"

app/assets/database/queries/asset_reference.py

@@ -266,9 +266,18 @@ def list_references_page(
     metadata_filter: dict | None = None,
     sort: str | None = None,
     order: str | None = None,
+    after_cursor_value: object | None = None,
+    after_cursor_id: str | None = None,
 ) -> tuple[list[AssetReference], dict[str, list[str]], int]:
     """List references with pagination, filtering, and sorting.
 
+    When ``after_cursor_value``/``after_cursor_id`` are supplied the query uses
+    keyset pagination — ``offset`` is ignored and a WHERE clause selects rows
+    strictly after the given ``(sort_col, id)`` position in the active sort
+    direction. The cursor value must already be typed for the column
+    (datetime for time sorts, int for size, str for name); the caller decodes
+    the opaque cursor string and resolves to the typed value.
+
     Returns (references, tag_map, total_count).
     """
     base = (
@@ -297,9 +306,31 @@ def list_references_page(
         "size": Asset.size_bytes,
     }
     sort_col = sort_map.get(sort, AssetReference.created_at)
-    sort_exp = sort_col.desc() if order == "desc" else sort_col.asc()
+    descending = order == "desc"
 
-    base = base.order_by(sort_exp).limit(limit).offset(offset)
+    # Keyset WHERE: (sort_col, id) strictly less-than / greater-than the cursor.
+    # Equivalent to: sort_col <op> v  OR  (sort_col = v AND id <op> cursor_id).
+    if after_cursor_value is not None and after_cursor_id is not None:
+        if descending:
+            keyset = sa.or_(
+                sort_col < after_cursor_value,
+                sa.and_(sort_col == after_cursor_value, AssetReference.id < after_cursor_id),
+            )
+        else:
+            keyset = sa.or_(
+                sort_col > after_cursor_value,
+                sa.and_(sort_col == after_cursor_value, AssetReference.id > after_cursor_id),
+            )
+        base = base.where(keyset)
+
+    # Secondary ORDER BY id (matching the primary direction) gives the keyset
+    # comparison a deterministic tiebreaker on duplicate sort_col values.
+    id_exp = AssetReference.id.desc() if descending else AssetReference.id.asc()
+    sort_exp = sort_col.desc() if descending else sort_col.asc()
+
+    base = base.order_by(sort_exp, id_exp).limit(limit)
+    if after_cursor_id is None:
+        base = base.offset(offset)
 
     count_stmt = (
         select(sa.func.count())

app/assets/database/queries/tags.py

@@ -55,13 +55,11 @@ def validate_tags_exist(session: Session, tags: list[str]) -> None:
         raise ValueError(f"Unknown tags: {missing}")
 
 
-def ensure_tags_exist(
-    session: Session, names: Iterable[str], tag_type: str = "user"
-) -> None:
+def ensure_tags_exist(session: Session, names: Iterable[str]) -> None:
     wanted = normalize_tags(list(names))
     if not wanted:
         return
-    rows = [{"name": n, "tag_type": tag_type} for n in list(dict.fromkeys(wanted))]
+    rows = [{"name": n} for n in list(dict.fromkeys(wanted))]
     ins = (
         sqlite.insert(Tag)
         .values(rows)
@@ -97,7 +95,7 @@ def set_reference_tags(
     to_remove = [t for t in current if t not in desired]
 
     if to_add:
-        ensure_tags_exist(session, to_add, tag_type="user")
+        ensure_tags_exist(session, to_add)
         session.add_all(
             [
                 AssetReferenceTag(
@@ -142,7 +140,7 @@ def add_tags_to_reference(
         return AddTagsResult(added=[], already_present=[], total_tags=total)
 
     if create_if_missing:
-        ensure_tags_exist(session, norm, tag_type="user")
+        ensure_tags_exist(session, norm)
 
     current = set(get_reference_tags(session, reference_id))
 
@@ -289,7 +287,6 @@ def list_tags_with_usage(
     q = (
         select(
             Tag.name,
-            Tag.tag_type,
             func.coalesce(counts_sq.c.cnt, 0).label("count"),
         )
         .select_from(Tag)
@@ -331,7 +328,7 @@ def list_tags_with_usage(
     rows = (session.execute(q.limit(limit).offset(offset))).all()
     total = (session.execute(total_q)).scalar_one()
 
-    rows_norm = [(name, ttype, int(count or 0)) for (name, ttype, count) in rows]
+    rows_norm = [(name, int(count or 0)) for (name, count) in rows]
     return rows_norm, int(total or 0)
 
 

app/assets/scanner.py

@@ -33,6 +33,7 @@ from app.assets.services.file_utils import (
     verify_file_unchanged,
 )
 from app.assets.services.hashing import HashCheckpoint, compute_blake3_hash
+from app.assets.services.image_dimensions import extract_image_dimensions
 from app.assets.services.metadata_extract import extract_file_metadata
 from app.assets.services.path_utils import (
     compute_relative_filename,
@@ -354,7 +355,7 @@ def insert_asset_specs(specs: list[SeedAssetSpec], tag_pool: set[str]) -> int:
         return 0
     with create_session() as sess:
         if tag_pool:
-            ensure_tags_exist(sess, tag_pool, tag_type="user")
+            ensure_tags_exist(sess, tag_pool)
         result = batch_insert_seed_assets(sess, specs=specs, owner_id="")
         sess.commit()
         return result.inserted_refs
@@ -506,6 +507,10 @@ def enrich_asset(
 
     if extract_metadata and metadata:
         system_metadata = metadata.to_user_metadata()
+        if mime_type and mime_type.startswith("image/"):
+            dims = extract_image_dimensions(file_path, mime_type=mime_type)
+            if dims:
+                system_metadata.update(dims)
         set_reference_system_metadata(session, reference_id, system_metadata)
 
     if full_hash:

app/assets/services/asset_management.py

@@ -1,8 +1,19 @@
 import contextlib
 import mimetypes
 import os
+from datetime import timezone
 from typing import Sequence
 
+from app.assets.services.cursor import (
+    CursorPayload,
+    InvalidCursorError,
+    decode_cursor,
+    decode_cursor_int,
+    decode_cursor_time,
+    encode_cursor,
+    encode_cursor_from_time,
+)
+
 
 from app.assets.database.models import Asset
 from app.assets.database.queries import (
@@ -149,6 +160,16 @@ def delete_asset_reference(
     owner_id: str,
     delete_content_if_orphan: bool = True,
 ) -> bool:
+    """Delete an asset reference.
+
+    With ``delete_content_if_orphan=False`` (a soft delete), the reference is
+    hidden and the underlying content is preserved. With ``True``, the content
+    is also removed once it becomes orphaned.
+
+    Note: the public DELETE /api/assets/{id} endpoint always soft-deletes
+    (passes ``False``); the orphan-reclamation path is intentionally
+    internal-only, retained for a future GC/admin caller.
+    """
     with create_session() as session:
         if not delete_content_if_orphan:
             # Soft delete: mark the reference as deleted but keep everything
@@ -242,6 +263,11 @@ def get_asset_by_hash(asset_hash: str) -> AssetData | None:
         return extract_asset_data(asset)
 
 
+# Sort fields that support cursor pagination. `last_access_time` is not
+# in this list — it falls back to offset/limit.
+_CURSOR_SORT_FIELDS = ("created_at", "updated_at", "name", "size")
+
+
 def list_assets_page(
     owner_id: str = "",
     include_tags: Sequence[str] | None = None,
@@ -252,7 +278,39 @@ def list_assets_page(
     offset: int = 0,
     sort: str = "created_at",
     order: str = "desc",
+    after: str | None = None,
 ) -> ListAssetsResult:
+    """List assets with optional cursor pagination.
+
+    When ``after`` is supplied it overrides ``offset``. The cursor's sort field
+    must match ``sort`` and be in the cursor-supported allowlist; mismatches
+    raise InvalidCursorError so the handler can map to 400 INVALID_CURSOR.
+    """
+    cursor_value: object | None = None
+    cursor_id: str | None = None
+    # Mint next_cursor on every page where the sort is cursor-supported, not
+    # only when the request itself arrived with a cursor. Otherwise a first
+    # request (no `after`) returns next_cursor=None and the client can never
+    # enter cursor mode.
+    mint_cursor = sort in _CURSOR_SORT_FIELDS
+
+    if after is not None:
+        if sort not in _CURSOR_SORT_FIELDS:
+            raise InvalidCursorError(
+                f"cursor pagination is not supported for sort={sort!r}"
+            )
+        payload = decode_cursor(after, _CURSOR_SORT_FIELDS, expected_order=order)
+        if payload.sort_field != sort:
+            raise InvalidCursorError(
+                f"cursor sort field {payload.sort_field!r} does not match request sort {sort!r}"
+            )
+        cursor_value, cursor_id = _resolve_cursor_value(payload), payload.id
+
+    # Over-fetch by one row so we can distinguish "exactly `limit` rows total
+    # remaining" from "more rows past this page" without a second query. Drop
+    # the sentinel before returning.
+    fetch_limit = limit + 1 if mint_cursor else limit
+
     with create_session() as session:
         refs, tag_map, total = list_references_page(
             session,
@@ -261,12 +319,22 @@ def list_assets_page(
             exclude_tags=exclude_tags,
             name_contains=name_contains,
             metadata_filter=metadata_filter,
-            limit=limit,
+            limit=fetch_limit,
             offset=offset,
             sort=sort,
             order=order,
+            after_cursor_value=cursor_value,
+            after_cursor_id=cursor_id,
         )
 
+        next_cursor: str | None = None
+        if mint_cursor and len(refs) > limit:
+            # There's at least one more row past this page — mint a cursor from
+            # the last row of the page (i.e. index `limit - 1`, since we
+            # over-fetched), and drop the sentinel.
+            next_cursor = _encode_next_cursor(refs[limit - 1], sort, order)
+            refs = refs[:limit]
+
         items: list[AssetSummaryData] = []
         for ref in refs:
             items.append(
@@ -277,7 +345,39 @@ def list_assets_page(
                 )
             )
 
-        return ListAssetsResult(items=items, total=total)
+        return ListAssetsResult(items=items, total=total, next_cursor=next_cursor)
+
+
+def _resolve_cursor_value(payload: CursorPayload) -> object:
+    """Map a decoded cursor payload to a column-typed Python value."""
+    if payload.sort_field in ("created_at", "updated_at"):
+        # DB stores naive UTC; strip tzinfo so the comparison binds against a
+        # `TIMESTAMP WITHOUT TIME ZONE` column without an offset shift.
+        return decode_cursor_time(payload).replace(tzinfo=None)
+    if payload.sort_field == "size":
+        return decode_cursor_int(payload)
+    return payload.value  # name, str-typed
+
+
+def _encode_next_cursor(ref, sort: str, order: str) -> str | None:
+    """Mint a cursor pointing at *ref* for the given sort dimension.
+
+    Returns None when the boundary row carries a NULL sort value (e.g. an asset
+    record whose size_bytes hasn't been backfilled). Continuing pagination
+    across a NULL boundary is undefined under keyset ordering — better to
+    truncate cleanly here than to mint a cursor that mis-positions.
+    """
+    if sort == "name":
+        return encode_cursor("name", ref.name, ref.id, order=order)
+    if sort == "size":
+        if ref.asset is None or ref.asset.size_bytes is None:
+            return None
+        return encode_cursor("size", str(ref.asset.size_bytes), ref.id, order=order)
+    # created_at / updated_at — DB datetimes are naive UTC; attach tz before encoding.
+    value = ref.created_at if sort == "created_at" else ref.updated_at
+    if value is None:
+        return None
+    return encode_cursor_from_time(sort, value.replace(tzinfo=timezone.utc), ref.id, order=order)
 
 
 def resolve_hash_to_path(

app/assets/services/cursor.py

@@ -0,0 +1,213 @@
+"""Opaque keyset-pagination cursor for /api/assets.
+
+Payload JSON uses short keys to keep the encoded length small:
+
+    {"s": <sort_field>, "v": <value>, "id": <id>, "o": <order>}
+
+The `o` key binds the cursor to the sort direction it was minted under,
+so replaying a `desc` cursor against an `asc` request fails with
+``INVALID_CURSOR`` rather than silently walking the wrong direction.
+`o` is mandatory on every payload — a cursor without it is rejected as
+malformed.
+
+Encoding is base64url with no padding. Cursors are opaque tokens: the
+payload format is internal to this server, and clients must treat a
+cursor as a black box handed back via `next_cursor`. No byte-level
+compatibility with any other implementation is required.
+
+Time values are serialized as Unix microseconds (UTC) — microsecond
+precision is sufficient to round-trip the timestamps stored by the
+database without rounding rows in the same millisecond bucket.
+"""
+from __future__ import annotations
+
+import base64
+import json
+from dataclasses import dataclass
+from datetime import datetime, timezone
+from typing import Iterable, Optional
+
+
+class InvalidCursorError(ValueError):
+    """Raised on a malformed, oversized, or unsupported-sort-field cursor.
+
+    Map to a 400 response with code ``INVALID_CURSOR`` at the handler.
+    """
+
+
+# Wire-format length caps. Cursors are user-controlled, so caps protect the
+# decode path from oversized allocations and downstream SQL predicates from
+# unbounded strings.
+#
+# MAX_CURSOR_VALUE_LENGTH is 512 to fit the `AssetReference.name` column max
+# (`String(512)`) — otherwise a long-named asset would mint a cursor the same
+# server then refuses on the next request.
+#
+# MAX_ENCODED_CURSOR_LENGTH is the decode-path guard, sized comfortably above
+# the largest cursor the per-field caps can produce. Worst case is value + id
+# at their caps with every character JSON-escaping to the six-byte `\uXXXX`
+# form (control characters), which is ~5.2 KB once base64url-encoded. At 8192
+# the encoder can never mint a cursor that exceeds it, so a freshly minted
+# cursor always decodes on the next request and there is no user-visible
+# "cursor too long" failure.
+MAX_ENCODED_CURSOR_LENGTH = 8192
+MAX_CURSOR_VALUE_LENGTH = 512
+MAX_CURSOR_ID_LENGTH = 128
+
+
+@dataclass(frozen=True)
+class CursorPayload:
+    sort_field: str
+    value: str
+    id: str
+    order: str
+
+
+_VALID_ORDERS = ("asc", "desc")
+
+
+def encode_cursor(sort_field: str, value: str, id: str, order: str = "desc") -> str:
+    """Encode a cursor payload as a base64url (no-padding) string.
+
+    `order` binds the cursor to the sort direction it was minted under so a
+    later request with a flipped `order` query parameter is rejected with
+    ``INVALID_CURSOR`` rather than silently walking the wrong direction.
+    """
+    if order not in _VALID_ORDERS:
+        raise InvalidCursorError(f"order must be one of {_VALID_ORDERS}, got {order!r}")
+    # Symmetric input validation: the encoder must reject anything the
+    # decoder rejects, or the same server will mint cursors it then 400s on
+    # the next request.
+    if not id:
+        raise InvalidCursorError("id must be non-empty")
+    if len(id) > MAX_CURSOR_ID_LENGTH:
+        raise InvalidCursorError("id exceeds maximum length")
+    if len(value) > MAX_CURSOR_VALUE_LENGTH:
+        raise InvalidCursorError("value exceeds maximum length")
+    payload = {"s": sort_field, "v": value, "id": id, "o": order}
+    raw = json.dumps(payload, separators=(",", ":"), ensure_ascii=False)
+    # No mint-time length guard is needed: the per-field caps above bound the
+    # encoded length well below MAX_ENCODED_CURSOR_LENGTH (see its definition),
+    # so the encoder can never produce a cursor the decode path would reject.
+    return base64.urlsafe_b64encode(raw.encode("utf-8")).rstrip(b"=").decode("ascii")
+
+
+def encode_cursor_from_time(sort_field: str, t: datetime, id: str, order: str = "desc") -> str:
+    """Encode a time-typed cursor at Unix microsecond precision.
+
+    Accepts an aware datetime (any timezone) and normalizes to UTC. Naive
+    datetimes are rejected so callers can't accidentally encode the local
+    wall-clock value of a UTC-stored timestamp.
+    """
+    if t.tzinfo is None:
+        raise ValueError("encode_cursor_from_time requires an aware datetime")
+    micros = _datetime_to_unix_micros(t.astimezone(timezone.utc))
+    return encode_cursor(sort_field, str(micros), id, order=order)
+
+
+def decode_cursor(
+    cursor: str,
+    allowed_sort_fields: Iterable[str],
+    expected_order: str | None = None,
+) -> CursorPayload:
+    """Parse an opaque cursor.
+
+    ``allowed_sort_fields`` is the endpoint's accepted sort-field list — a
+    cursor carrying a field outside this set is rejected so a cursor minted
+    for one column can't be replayed against another (e.g. a ``created_at``
+    timestamp string compared against a ``name`` column).
+
+    ``expected_order`` (``"asc"``/``"desc"``), when supplied, must match the
+    payload's ``o`` field. ``o`` is required on every payload; a cursor
+    missing it is rejected as malformed.
+
+    Passing no allowed fields rejects every cursor.
+    """
+    if len(cursor) > MAX_ENCODED_CURSOR_LENGTH:
+        raise InvalidCursorError("cursor exceeds maximum length")
+
+    try:
+        # urlsafe_b64decode requires correct padding; we strip on encode, so
+        # restore the trailing '=' pad here.
+        padding = "=" * (-len(cursor) % 4)
+        raw = base64.urlsafe_b64decode(cursor + padding)
+    except (ValueError, base64.binascii.Error) as e:
+        raise InvalidCursorError(f"encoding: {e}") from e
+
+    try:
+        decoded = json.loads(raw)
+    except (json.JSONDecodeError, UnicodeDecodeError) as e:
+        raise InvalidCursorError(f"payload: {e}") from e
+
+    if not isinstance(decoded, dict):
+        raise InvalidCursorError("payload: expected object")
+
+    sort_field = decoded.get("s")
+    value = decoded.get("v")
+    id = decoded.get("id")
+    order = decoded.get("o")
+
+    if not isinstance(sort_field, str) or not isinstance(value, str) or not isinstance(id, str):
+        raise InvalidCursorError("payload: missing or non-string s/v/id")
+
+    if id == "":
+        raise InvalidCursorError("missing id")
+    if len(id) > MAX_CURSOR_ID_LENGTH:
+        raise InvalidCursorError("id exceeds maximum length")
+    if len(value) > MAX_CURSOR_VALUE_LENGTH:
+        raise InvalidCursorError("value exceeds maximum length")
+
+    if sort_field not in allowed_sort_fields:
+        raise InvalidCursorError(f"unsupported sort field {sort_field!r}")
+
+    if not isinstance(order, str):
+        raise InvalidCursorError("missing or non-string o")
+    if order not in _VALID_ORDERS:
+        raise InvalidCursorError(f"unsupported order {order!r}")
+    if expected_order is not None and order != expected_order:
+        raise InvalidCursorError(
+            f"cursor order {order!r} does not match request order {expected_order!r}"
+        )
+
+    return CursorPayload(sort_field=sort_field, value=value, id=id, order=order)
+
+
+def decode_cursor_time(payload: Optional[CursorPayload]) -> datetime:
+    """Parse a time-typed cursor value as Unix microseconds, returning UTC."""
+    if payload is None:
+        raise InvalidCursorError("nil cursor payload")
+    try:
+        micros = int(payload.value)
+    except ValueError as e:
+        raise InvalidCursorError(f"value is not a valid timestamp: {e}") from e
+    try:
+        return _unix_micros_to_datetime(micros)
+    except (OverflowError, OSError, ValueError) as e:
+        # Crafted out-of-range microseconds (e.g. > datetime.MAX_YEAR) blow up
+        # in fromtimestamp / datetime construction. Map to 400, not 500.
+        raise InvalidCursorError(f"value is out of representable range: {e}") from e
+
+
+def decode_cursor_int(payload: Optional[CursorPayload]) -> int:
+    """Parse a cursor value as a base-10 integer."""
+    if payload is None:
+        raise InvalidCursorError("nil cursor payload")
+    try:
+        return int(payload.value)
+    except ValueError as e:
+        raise InvalidCursorError(f"value is not a valid integer: {e}") from e
+
+
+_EPOCH = datetime(1970, 1, 1, tzinfo=timezone.utc)
+
+
+def _datetime_to_unix_micros(t: datetime) -> int:
+    """Convert an aware UTC datetime to Unix microseconds (integer math)."""
+    delta = t - _EPOCH
+    return (delta.days * 86_400 + delta.seconds) * 1_000_000 + delta.microseconds
+
+
+def _unix_micros_to_datetime(micros: int) -> datetime:
+    """Convert Unix microseconds to a UTC datetime, preserving precision."""
+    seconds, micro_remainder = divmod(micros, 1_000_000)
+    return datetime.fromtimestamp(seconds, tz=timezone.utc).replace(microsecond=micro_remainder)

app/assets/services/image_dimensions.py

@@ -0,0 +1,63 @@
+"""Image dimension extraction for asset ingest.
+
+Reads only the image header via Pillow to capture width/height cheaply,
+without a full pixel decode. Returns a metadata dict suitable for merging
+into ``AssetReference.system_metadata``.
+"""
+from __future__ import annotations
+
+import logging
+from typing import Any
+
+logger = logging.getLogger(__name__)
+
+
+def extract_image_dimensions(
+    file_path: str, mime_type: str | None = None
+) -> dict[str, Any] | None:
+    """Extract image dimensions for the file at ``file_path``.
+
+    Args:
+        file_path: Absolute path to a file on disk.
+        mime_type: Optional MIME type hint. When provided and not prefixed
+            with ``image/``, extraction is skipped without touching the file.
+
+    Returns:
+        ``{"kind": "image", "width": W, "height": H}`` when the file is a
+        recognizable image with positive dimensions, otherwise ``None``.
+
+    The dict shape is intended to be merged into ``system_metadata`` so the
+    asset response surfaces ``metadata.kind`` plus dimension fields for image
+    assets. Forward-compatible: future media kinds (e.g. ``"video"`` with
+    duration/fps) can extend this shape without schema changes.
+    """
+    if mime_type is not None and not mime_type.startswith("image/"):
+        return None
+
+    try:
+        from PIL import Image, UnidentifiedImageError
+    except ImportError:
+        logger.debug(
+            "Pillow not available; skipping image dimension extraction for %s",
+            file_path,
+        )
+        return None
+
+    try:
+        with Image.open(file_path) as img:
+            width, height = img.size
+    except (OSError, UnidentifiedImageError, ValueError) as exc:
+        logger.debug(
+            "Failed to read image dimensions from %s: %s", file_path, exc
+        )
+        return None
+
+    if (
+        not isinstance(width, int)
+        or not isinstance(height, int)
+        or width <= 0
+        or height <= 0
+    ):
+        return None
+
+    return {"kind": "image", "width": width, "height": height}

app/assets/services/ingest.py

@@ -17,9 +17,11 @@ from app.assets.database.queries import (
     get_reference_by_file_path,
     get_reference_tags,
     get_or_create_reference,
+    list_references_by_asset_id,
     reference_exists,
     remove_missing_tag_for_asset_id,
     set_reference_metadata,
+    set_reference_system_metadata,
     set_reference_tags,
     update_asset_hash_and_mime,
     upsert_asset,
@@ -29,6 +31,7 @@ from app.assets.database.queries import (
 from app.assets.helpers import get_utc_now, normalize_tags
 from app.assets.services.bulk_ingest import batch_insert_seed_assets
 from app.assets.services.file_utils import get_size_and_mtime_ns
+from app.assets.services.image_dimensions import extract_image_dimensions
 from app.assets.services.path_utils import (
     compute_relative_filename,
     get_name_and_tags_from_asset_path,
@@ -118,6 +121,14 @@ def _ingest_file_from_path(
                 user_metadata=user_metadata,
             )
 
+            _maybe_store_image_dimensions(
+                session,
+                reference_id=reference_id,
+                file_path=locator,
+                mime_type=mime_type,
+                current_system_metadata=ref.system_metadata,
+            )
+
         try:
             remove_missing_tag_for_asset_id(session, asset_id=asset.id)
         except Exception:
@@ -288,6 +299,13 @@ def _register_existing_asset(
                 user_metadata=new_meta,
             )
 
+        _backfill_image_dimensions_from_siblings(
+            session,
+            asset_id=asset.id,
+            new_reference_id=ref.id,
+            current_system_metadata=ref.system_metadata,
+        )
+
         if tags is not None:
             set_reference_tags(
                 session,
@@ -334,6 +352,87 @@ def _update_metadata_with_filename(
         )
 
 
+_IMAGE_DIMENSION_KEYS = ("kind", "width", "height")
+
+
+def _maybe_store_image_dimensions(
+    session: Session,
+    reference_id: str,
+    file_path: str,
+    mime_type: str | None,
+    current_system_metadata: dict | None,
+) -> None:
+    """Populate ``kind``/``width``/``height`` on system_metadata for image refs.
+
+    Non-image MIME types are a no-op. Pre-existing keys (e.g. enricher-written
+    safetensors metadata, download provenance) are preserved by merge.
+    """
+    if not mime_type or not mime_type.startswith("image/"):
+        return
+
+    dims = extract_image_dimensions(file_path, mime_type=mime_type)
+    if not dims:
+        return
+
+    current = current_system_metadata or {}
+    merged = dict(current)
+    merged.update(dims)
+    if merged != current:
+        set_reference_system_metadata(
+            session,
+            reference_id=reference_id,
+            system_metadata=merged,
+        )
+
+
+def _backfill_image_dimensions_from_siblings(
+    session: Session,
+    asset_id: str,
+    new_reference_id: str,
+    current_system_metadata: dict | None,
+) -> None:
+    """Copy image dimension keys from any sibling reference of the same asset.
+
+    The from-hash path doesn't read the file bytes, so dimensions can't be
+    extracted there directly. When another reference of the same asset already
+    carries image dimensions, copy them onto the new reference so consumers
+    see consistent metadata regardless of how the asset was registered.
+
+    Best-effort: missing siblings, non-image siblings, or absent dimension
+    keys leave the target reference unchanged.
+    """
+    current = current_system_metadata or {}
+    if current.get("kind") == "image" and "width" in current and "height" in current:
+        return
+
+    for sibling in list_references_by_asset_id(session, asset_id):
+        if sibling.id == new_reference_id:
+            continue
+        meta = sibling.system_metadata or {}
+        if meta.get("kind") != "image":
+            continue
+        width = meta.get("width")
+        height = meta.get("height")
+        if (
+            type(width) is not int
+            or type(height) is not int
+            or width <= 0
+            or height <= 0
+        ):
+            continue
+        merged = dict(current)
+        merged["kind"] = "image"
+        merged["width"] = width
+        merged["height"] = height
+        if merged != current:
+            set_reference_system_metadata(
+                session,
+                reference_id=new_reference_id,
+                system_metadata=merged,
+            )
+        return
+
+
 def _sanitize_filename(name: str | None, fallback: str) -> str:
     n = os.path.basename((name or "").strip() or fallback)
     return n if n else fallback

app/assets/services/schemas.py

@@ -56,7 +56,6 @@ class IngestResult:
 
 class TagUsage(NamedTuple):
     name: str
-    tag_type: str
     count: int
 
 
@@ -71,6 +70,7 @@ class AssetSummaryData:
 class ListAssetsResult:
     items: list[AssetSummaryData]
     total: int
+    next_cursor: str | None = None
 
 
 @dataclass(frozen=True)

app/assets/services/tagging.py

@@ -75,7 +75,7 @@ def list_tags(
             owner_id=owner_id,
         )
 
-    return [TagUsage(name, tag_type, count) for name, tag_type, count in rows], total
+    return [TagUsage(name, count) for name, count in rows], total
 
 
 def list_tag_histogram(

blueprints/Image Depth Estimation (Depth Anything 3).json

@@ -0,0 +1,569 @@
+{
+  "revision": 0,
+  "last_node_id": 89,
+  "last_link_id": 0,
+  "nodes": [
+    {
+      "id": 89,
+      "type": "85e595bd-af9e-40ee-85c5-b98bb15da47a",
+      "pos": [
+        320,
+        520
+      ],
+      "size": [
+        400,
+        360
+      ],
+      "flags": {},
+      "order": 3,
+      "mode": 0,
+      "inputs": [
+        {
+          "localized_name": "image",
+          "name": "image",
+          "type": "IMAGE",
+          "link": null
+        },
+        {
+          "name": "resolution",
+          "type": "INT",
+          "widget": {
+            "name": "resolution"
+          },
+          "link": null
+        },
+        {
+          "name": "resize_method",
+          "type": "COMBO",
+          "widget": {
+            "name": "resize_method"
+          },
+          "link": null
+        },
+        {
+          "label": "output_type",
+          "name": "output",
+          "type": "COMFY_DYNAMICCOMBO_V3",
+          "widget": {
+            "name": "output"
+          },
+          "link": null
+        },
+        {
+          "label": "output_normalization",
+          "name": "output.normalization",
+          "type": "COMBO",
+          "widget": {
+            "name": "output.normalization"
+          },
+          "link": null
+        },
+        {
+          "label": "apply_sky_clip",
+          "name": "output.apply_sky_clip",
+          "type": "BOOLEAN",
+          "widget": {
+            "name": "output.apply_sky_clip"
+          },
+          "link": null
+        },
+        {
+          "name": "model_name",
+          "type": "COMBO",
+          "widget": {
+            "name": "model_name"
+          },
+          "link": null
+        }
+      ],
+      "outputs": [
+        {
+          "localized_name": "IMAGE",
+          "name": "IMAGE",
+          "type": "IMAGE",
+          "links": []
+        }
+      ],
+      "properties": {
+        "proxyWidgets": [
+          [
+            "87",
+            "resolution"
+          ],
+          [
+            "87",
+            "resize_method"
+          ],
+          [
+            "86",
+            "output"
+          ],
+          [
+            "86",
+            "output.normalization"
+          ],
+          [
+            "86",
+            "output.apply_sky_clip"
+          ],
+          [
+            "88",
+            "model_name"
+          ]
+        ],
+        "cnr_id": "comfy-core",
+        "ver": "0.24.0"
+      },
+      "widgets_values": [],
+      "title": "Image Depth Estimation (Depth Anything 3)"
+    }
+  ],
+  "links": [],
+  "version": 0.4,
+  "definitions": {
+    "subgraphs": [
+      {
+        "id": "85e595bd-af9e-40ee-85c5-b98bb15da47a",
+        "version": 1,
+        "state": {
+          "lastGroupId": 4,
+          "lastNodeId": 89,
+          "lastLinkId": 109,
+          "lastRerouteId": 0
+        },
+        "revision": 2,
+        "config": {},
+        "name": "Image Depth Estimation (Depth Anything 3)",
+        "inputNode": {
+          "id": -10,
+          "bounding": [
+            400,
+            90,
+            166.998046875,
+            188
+          ]
+        },
+        "outputNode": {
+          "id": -20,
+          "bounding": [
+            1250,
+            146,
+            128,
+            68
+          ]
+        },
+        "inputs": [
+          {
+            "id": "43cf3118-495a-487d-8eb3-a17c7e92f64f",
+            "name": "image",
+            "type": "IMAGE",
+            "linkIds": [
+              19
+            ],
+            "localized_name": "image",
+            "pos": [
+              542.998046875,
+              114
+            ]
+          },
+          {
+            "id": "1089a0a1-6db1-45a8-84b0-0bfdc2ed920a",
+            "name": "resolution",
+            "type": "INT",
+            "linkIds": [
+              22
+            ],
+            "pos": [
+              542.998046875,
+              134
+            ]
+          },
+          {
+            "id": "25fb64ac-26d5-466d-995b-6d51b9afa2c4",
+            "name": "resize_method",
+            "type": "COMBO",
+            "linkIds": [
+              23
+            ],
+            "pos": [
+              542.998046875,
+              154
+            ]
+          },
+          {
+            "id": "8acafb7c-6c8b-46b3-9d74-c563498a3af1",
+            "name": "output",
+            "type": "COMFY_DYNAMICCOMBO_V3",
+            "linkIds": [
+              24
+            ],
+            "label": "output_type",
+            "pos": [
+              542.998046875,
+              174
+            ]
+          },
+          {
+            "id": "1da5009b-4648-43e8-a257-16426630cf22",
+            "name": "output.normalization",
+            "type": "COMBO",
+            "linkIds": [
+              25
+            ],
+            "label": "output_normalization",
+            "pos": [
+              542.998046875,
+              194
+            ]
+          },
+          {
+            "id": "fd7edb33-5fb1-4538-a411-26e5039a9321",
+            "name": "output.apply_sky_clip",
+            "type": "BOOLEAN",
+            "linkIds": [
+              26
+            ],
+            "label": "apply_sky_clip",
+            "pos": [
+              542.998046875,
+              214
+            ]
+          },
+          {
+            "id": "b5be4c8a-b833-4f1e-8c94-3ed1dd722190",
+            "name": "model_name",
+            "type": "COMBO",
+            "linkIds": [
+              106
+            ],
+            "pos": [
+              542.998046875,
+              234
+            ]
+          }
+        ],
+        "outputs": [
+          {
+            "id": "478ab537-63bc-4d74-a9f0-c975f550880f",
+            "name": "IMAGE",
+            "type": "IMAGE",
+            "linkIds": [
+              7
+            ],
+            "localized_name": "IMAGE",
+            "pos": [
+              1274,
+              170
+            ]
+          }
+        ],
+        "widgets": [],
+        "nodes": [
+          {
+            "id": 86,
+            "type": "DA3Render",
+            "pos": [
+              800,
+              310
+            ],
+            "size": [
+              380,
+              130
+            ],
+            "flags": {},
+            "order": 0,
+            "mode": 0,
+            "inputs": [
+              {
+                "localized_name": "da3_geometry",
+                "name": "da3_geometry",
+                "type": "DA3_GEOMETRY",
+                "link": 12
+              },
+              {
+                "localized_name": "output",
+                "name": "output",
+                "type": "COMFY_DYNAMICCOMBO_V3",
+                "widget": {
+                  "name": "output"
+                },
+                "link": 24
+              },
+              {
+                "localized_name": "output.normalization",
+                "name": "output.normalization",
+                "type": "COMBO",
+                "widget": {
+                  "name": "output.normalization"
+                },
+                "link": 25
+              },
+              {
+                "localized_name": "output.apply_sky_clip",
+                "name": "output.apply_sky_clip",
+                "type": "BOOLEAN",
+                "widget": {
+                  "name": "output.apply_sky_clip"
+                },
+                "link": 26
+              },
+              {
+                "name": "geometry",
+                "type": "DA3_GEOMETRY",
+                "link": null
+              }
+            ],
+            "outputs": [
+              {
+                "localized_name": "IMAGE",
+                "name": "IMAGE",
+                "type": "IMAGE",
+                "slot_index": 0,
+                "links": [
+                  7
+                ]
+              }
+            ],
+            "properties": {
+              "Node name for S&R": "DA3Render",
+              "cnr_id": "comfy-core",
+              "ver": "0.19.0"
+            },
+            "widgets_values": [
+              "depth",
+              "v2_style",
+              false
+            ]
+          },
+          {
+            "id": 87,
+            "type": "DA3Inference",
+            "pos": [
+              800,
+              50
+            ],
+            "size": [
+              390,
+              130
+            ],
+            "flags": {},
+            "order": 1,
+            "mode": 0,
+            "inputs": [
+              {
+                "localized_name": "da3_model",
+                "name": "da3_model",
+                "type": "DA3_MODEL",
+                "link": 107
+              },
+              {
+                "localized_name": "image",
+                "name": "image",
+                "type": "IMAGE",
+                "link": 19
+              },
+              {
+                "localized_name": "resolution",
+                "name": "resolution",
+                "type": "INT",
+                "widget": {
+                  "name": "resolution"
+                },
+                "link": 22
+              },
+              {
+                "localized_name": "resize_method",
+                "name": "resize_method",
+                "type": "COMBO",
+                "widget": {
+                  "name": "resize_method"
+                },
+                "link": 23
+              },
+              {
+                "localized_name": "mode",
+                "name": "mode",
+                "type": "COMFY_DYNAMICCOMBO_V3",
+                "widget": {
+                  "name": "mode"
+                },
+                "link": null
+              }
+            ],
+            "outputs": [
+              {
+                "localized_name": "da3_geometry",
+                "name": "da3_geometry",
+                "type": "DA3_GEOMETRY",
+                "slot_index": 0,
+                "links": [
+                  12
+                ]
+              }
+            ],
+            "properties": {
+              "Node name for S&R": "DA3Inference",
+              "cnr_id": "comfy-core",
+              "ver": "0.19.0"
+            },
+            "widgets_values": [
+              504,
+              "upper_bound_resize",
+              "mono"
+            ]
+          },
+          {
+            "id": 88,
+            "type": "LoadDA3Model",
+            "pos": [
+              810,
+              -160
+            ],
+            "size": [
+              400,
+              140
+            ],
+            "flags": {},
+            "order": 2,
+            "mode": 0,
+            "inputs": [
+              {
+                "localized_name": "model_name",
+                "name": "model_name",
+                "type": "COMBO",
+                "widget": {
+                  "name": "model_name"
+                },
+                "link": 106
+              },
+              {
+                "localized_name": "weight_dtype",
+                "name": "weight_dtype",
+                "type": "COMBO",
+                "widget": {
+                  "name": "weight_dtype"
+                },
+                "link": null
+              }
+            ],
+            "outputs": [
+              {
+                "localized_name": "DA3_MODEL",
+                "name": "DA3_MODEL",
+                "type": "DA3_MODEL",
+                "links": [
+                  107
+                ]
+              }
+            ],
+            "properties": {
+              "Node name for S&R": "LoadDA3Model",
+              "cnr_id": "comfy-core",
+              "ver": "0.24.0",
+              "models": [
+                {
+                  "name": "depth_anything_3_mono_large.safetensors",
+                  "url": "https://huggingface.co/Comfy-Org/Depth-Anything-3/resolve/main/geometry_estimation/depth_anything_3_mono_large.safetensors",
+                  "directory": "geometry_estimation"
+                }
+              ]
+            },
+            "widgets_values": [
+              "depth_anything_3_mono_large.safetensors",
+              "default"
+            ]
+          }
+        ],
+        "groups": [],
+        "links": [
+          {
+            "id": 12,
+            "origin_id": 87,
+            "origin_slot": 0,
+            "target_id": 86,
+            "target_slot": 0,
+            "type": "DA3_GEOMETRY"
+          },
+          {
+            "id": 19,
+            "origin_id": -10,
+            "origin_slot": 0,
+            "target_id": 87,
+            "target_slot": 1,
+            "type": "IMAGE"
+          },
+          {
+            "id": 7,
+            "origin_id": 86,
+            "origin_slot": 0,
+            "target_id": -20,
+            "target_slot": 0,
+            "type": "IMAGE"
+          },
+          {
+            "id": 22,
+            "origin_id": -10,
+            "origin_slot": 1,
+            "target_id": 87,
+            "target_slot": 2,
+            "type": "INT"
+          },
+          {
+            "id": 23,
+            "origin_id": -10,
+            "origin_slot": 2,
+            "target_id": 87,
+            "target_slot": 3,
+            "type": "COMBO"
+          },
+          {
+            "id": 24,
+            "origin_id": -10,
+            "origin_slot": 3,
+            "target_id": 86,
+            "target_slot": 1,
+            "type": "COMFY_DYNAMICCOMBO_V3"
+          },
+          {
+            "id": 25,
+            "origin_id": -10,
+            "origin_slot": 4,
+            "target_id": 86,
+            "target_slot": 2,
+            "type": "COMBO"
+          },
+          {
+            "id": 26,
+            "origin_id": -10,
+            "origin_slot": 5,
+            "target_id": 86,
+            "target_slot": 3,
+            "type": "BOOLEAN"
+          },
+          {
+            "id": 106,
+            "origin_id": -10,
+            "origin_slot": 6,
+            "target_id": 88,
+            "target_slot": 0,
+            "type": "COMBO"
+          },
+          {
+            "id": 107,
+            "origin_id": 88,
+            "origin_slot": 0,
+            "target_id": 87,
+            "target_slot": 0,
+            "type": "DA3_MODEL"
+          }
+        ],
+        "extra": {},
+        "category": "Conditioning & Preprocessors/Depth",
+        "description": "This subgraph takes an input image and produces a depth map using the Depth Anything 3 model, which recovers spatially consistent geometry from any number of views. It is ideal for single or multi-view images, videos, and 3D scenes where accurate depth estimation is needed for tasks like SLAM, novel view synthesis, or spatial perception. The model uses a plain transformer backbone and supports both monocular and multi-view inputs without."
+      }
+    ]
+  },
+  "extra": {
+    "BlueprintDescription": "This subgraph takes an input image and produces a depth map using the Depth Anything 3 model, which recovers spatially consistent geometry from any number of views. It is ideal for single or multi-view images, videos, and 3D scenes where accurate depth estimation is needed for tasks like SLAM, novel view synthesis, or spatial perception. The model uses a plain transformer backbone and supports both monocular and multi-view inputs without."
+  }
+}

blueprints/Text to Image (Anima).json

@@ -1077,9 +1077,12 @@
           }
         ],
         "extra": {},
-        "category": "Image generation and editing/Text to image"
+        "category": "Image generation and editing/Text to image",
+        "description": "This subgraph converts text prompts into non-photorealistic illustrations using a 2-billion-parameter model optimized for anime and artistic styles. It is ideal for generating concept art, character designs, or stylized illustrations where photorealism is not required. The model excels with anime and artistic content but performs poorly on realistic subjects."
       }
     ]
   },
-  "extra": {}
+  "extra": {
+    "BlueprintDescription": "This subgraph converts text prompts into non-photorealistic illustrations using a 2-billion-parameter model optimized for anime and artistic styles. It is ideal for generating concept art, character designs, or stylized illustrations where photorealism is not required. The model excels with anime and artistic content but performs poorly on realistic subjects."
+  }
 }

blueprints/Video Depth Estimation (Depth Anything 3).json

@@ -0,0 +1,825 @@
+{
+  "revision": 0,
+  "last_node_id": 97,
+  "last_link_id": 0,
+  "nodes": [
+    {
+      "id": 97,
+      "type": "253ec5ca-8333-4ddf-a036-9fc0923651b9",
+      "pos": [
+        410,
+        500
+      ],
+      "size": [
+        400,
+        400
+      ],
+      "flags": {},
+      "order": 3,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "video",
+          "type": "VIDEO",
+          "link": null
+        },
+        {
+          "name": "start_time",
+          "type": "FLOAT",
+          "widget": {
+            "name": "start_time"
+          },
+          "link": null
+        },
+        {
+          "name": "duration",
+          "type": "FLOAT",
+          "widget": {
+            "name": "duration"
+          },
+          "link": null
+        },
+        {
+          "name": "resolution",
+          "type": "INT",
+          "widget": {
+            "name": "resolution"
+          },
+          "link": null
+        },
+        {
+          "name": "resize_method",
+          "type": "COMBO",
+          "widget": {
+            "name": "resize_method"
+          },
+          "link": null
+        },
+        {
+          "label": "output_type",
+          "name": "output",
+          "type": "COMFY_DYNAMICCOMBO_V3",
+          "widget": {
+            "name": "output"
+          },
+          "link": null
+        },
+        {
+          "label": "normalization",
+          "name": "output.normalization",
+          "type": "COMBO",
+          "widget": {
+            "name": "output.normalization"
+          },
+          "link": null
+        },
+        {
+          "name": "output.apply_sky_clip",
+          "type": "BOOLEAN",
+          "widget": {
+            "name": "output.apply_sky_clip"
+          },
+          "link": null
+        },
+        {
+          "name": "model_name",
+          "type": "COMBO",
+          "widget": {
+            "name": "model_name"
+          },
+          "link": null
+        }
+      ],
+      "outputs": [
+        {
+          "localized_name": "IMAGE",
+          "name": "IMAGE",
+          "type": "IMAGE",
+          "links": []
+        },
+        {
+          "name": "audio",
+          "type": "AUDIO",
+          "links": []
+        },
+        {
+          "name": "fps",
+          "type": "FLOAT",
+          "links": []
+        }
+      ],
+      "properties": {
+        "proxyWidgets": [
+          [
+            "96",
+            "start_time"
+          ],
+          [
+            "96",
+            "duration"
+          ],
+          [
+            "93",
+            "resolution"
+          ],
+          [
+            "93",
+            "resize_method"
+          ],
+          [
+            "92",
+            "output"
+          ],
+          [
+            "92",
+            "output.normalization"
+          ],
+          [
+            "92",
+            "output.apply_sky_clip"
+          ],
+          [
+            "94",
+            "model_name"
+          ]
+        ],
+        "cnr_id": "comfy-core",
+        "ver": "0.24.0"
+      },
+      "widgets_values": [],
+      "title": "Video Depth Estimation (Depth Anything 3)"
+    }
+  ],
+  "links": [],
+  "version": 0.4,
+  "definitions": {
+    "subgraphs": [
+      {
+        "id": "253ec5ca-8333-4ddf-a036-9fc0923651b9",
+        "version": 1,
+        "state": {
+          "lastGroupId": 4,
+          "lastNodeId": 97,
+          "lastLinkId": 129,
+          "lastRerouteId": 0
+        },
+        "revision": 2,
+        "config": {},
+        "name": "Video Depth Estimation (Depth Anything 3)",
+        "inputNode": {
+          "id": -10,
+          "bounding": [
+            -230,
+            130,
+            167.912109375,
+            228
+          ]
+        },
+        "outputNode": {
+          "id": -20,
+          "bounding": [
+            1520,
+            140,
+            128,
+            108
+          ]
+        },
+        "inputs": [
+          {
+            "id": "698c28c6-cf92-4039-8b39-f3062868ea7c",
+            "name": "video",
+            "type": "VIDEO",
+            "linkIds": [
+              119
+            ],
+            "pos": [
+              -86.087890625,
+              154
+            ]
+          },
+          {
+            "id": "97a1f63e-1585-4a40-9dec-e2700120d84a",
+            "name": "start_time",
+            "type": "FLOAT",
+            "linkIds": [
+              121
+            ],
+            "pos": [
+              -86.087890625,
+              174
+            ]
+          },
+          {
+            "id": "4dbbd3b3-c5ee-4a56-a0d3-3268d3b2fd64",
+            "name": "duration",
+            "type": "FLOAT",
+            "linkIds": [
+              122
+            ],
+            "pos": [
+              -86.087890625,
+              194
+            ]
+          },
+          {
+            "id": "16f55101-f99d-4c0c-bebf-c3b31c54f13e",
+            "name": "resolution",
+            "type": "INT",
+            "linkIds": [
+              124
+            ],
+            "pos": [
+              -86.087890625,
+              214
+            ]
+          },
+          {
+            "id": "d9cd7693-4bb3-4ed7-9a75-276b997abcd9",
+            "name": "resize_method",
+            "type": "COMBO",
+            "linkIds": [
+              125
+            ],
+            "pos": [
+              -86.087890625,
+              234
+            ]
+          },
+          {
+            "id": "a6e90532-323b-462e-ba9c-1672384d5b31",
+            "name": "output",
+            "type": "COMFY_DYNAMICCOMBO_V3",
+            "linkIds": [
+              126
+            ],
+            "label": "output_type",
+            "pos": [
+              -86.087890625,
+              254
+            ]
+          },
+          {
+            "id": "69e6aeef-437d-4fde-b2fc-d5ab9369238d",
+            "name": "output.normalization",
+            "type": "COMBO",
+            "linkIds": [
+              127
+            ],
+            "label": "normalization",
+            "pos": [
+              -86.087890625,
+              274
+            ]
+          },
+          {
+            "id": "73206f72-f89a-4698-885e-5d9277df2998",
+            "name": "output.apply_sky_clip",
+            "type": "BOOLEAN",
+            "linkIds": [
+              128
+            ],
+            "pos": [
+              -86.087890625,
+              294
+            ]
+          },
+          {
+            "id": "dddbc7fc-9431-448a-9ed3-9aa62404288b",
+            "name": "model_name",
+            "type": "COMBO",
+            "linkIds": [
+              129
+            ],
+            "pos": [
+              -86.087890625,
+              314
+            ]
+          }
+        ],
+        "outputs": [
+          {
+            "id": "478ab537-63bc-4d74-a9f0-c975f550880f",
+            "name": "IMAGE",
+            "type": "IMAGE",
+            "linkIds": [
+              7
+            ],
+            "localized_name": "IMAGE",
+            "pos": [
+              1544,
+              164
+            ]
+          },
+          {
+            "id": "cdaf037e-79bc-4a94-b06c-0fd32e76f615",
+            "name": "audio",
+            "type": "AUDIO",
+            "linkIds": [
+              112
+            ],
+            "pos": [
+              1544,
+              184
+            ]
+          },
+          {
+            "id": "4c0e5484-d193-49c7-b107-92619628880a",
+            "name": "fps",
+            "type": "FLOAT",
+            "linkIds": [
+              113
+            ],
+            "pos": [
+              1544,
+              204
+            ]
+          }
+        ],
+        "widgets": [],
+        "nodes": [
+          {
+            "id": 92,
+            "type": "DA3Render",
+            "pos": [
+              740,
+              230
+            ],
+            "size": [
+              380,
+              130
+            ],
+            "flags": {},
+            "order": 0,
+            "mode": 0,
+            "inputs": [
+              {
+                "localized_name": "da3_geometry",
+                "name": "da3_geometry",
+                "type": "DA3_GEOMETRY",
+                "link": 12
+              },
+              {
+                "localized_name": "output",
+                "name": "output",
+                "type": "COMFY_DYNAMICCOMBO_V3",
+                "widget": {
+                  "name": "output"
+                },
+                "link": 126
+              },
+              {
+                "localized_name": "output.normalization",
+                "name": "output.normalization",
+                "type": "COMBO",
+                "widget": {
+                  "name": "output.normalization"
+                },
+                "link": 127
+              },
+              {
+                "localized_name": "output.apply_sky_clip",
+                "name": "output.apply_sky_clip",
+                "type": "BOOLEAN",
+                "widget": {
+                  "name": "output.apply_sky_clip"
+                },
+                "link": 128
+              },
+              {
+                "name": "geometry",
+                "type": "DA3_GEOMETRY",
+                "link": null
+              }
+            ],
+            "outputs": [
+              {
+                "localized_name": "IMAGE",
+                "name": "IMAGE",
+                "type": "IMAGE",
+                "slot_index": 0,
+                "links": [
+                  7
+                ]
+              }
+            ],
+            "properties": {
+              "Node name for S&R": "DA3Render",
+              "cnr_id": "comfy-core",
+              "ver": "0.19.0"
+            },
+            "widgets_values": [
+              "depth",
+              "v2_style",
+              false
+            ]
+          },
+          {
+            "id": 93,
+            "type": "DA3Inference",
+            "pos": [
+              740,
+              -30
+            ],
+            "size": [
+              390,
+              130
+            ],
+            "flags": {},
+            "order": 1,
+            "mode": 0,
+            "inputs": [
+              {
+                "localized_name": "da3_model",
+                "name": "da3_model",
+                "type": "DA3_MODEL",
+                "link": 107
+              },
+              {
+                "localized_name": "image",
+                "name": "image",
+                "type": "IMAGE",
+                "link": 111
+              },
+              {
+                "localized_name": "resolution",
+                "name": "resolution",
+                "type": "INT",
+                "widget": {
+                  "name": "resolution"
+                },
+                "link": 124
+              },
+              {
+                "localized_name": "resize_method",
+                "name": "resize_method",
+                "type": "COMBO",
+                "widget": {
+                  "name": "resize_method"
+                },
+                "link": 125
+              },
+              {
+                "localized_name": "mode",
+                "name": "mode",
+                "type": "COMFY_DYNAMICCOMBO_V3",
+                "widget": {
+                  "name": "mode"
+                },
+                "link": null
+              }
+            ],
+            "outputs": [
+              {
+                "localized_name": "da3_geometry",
+                "name": "da3_geometry",
+                "type": "DA3_GEOMETRY",
+                "slot_index": 0,
+                "links": [
+                  12
+                ]
+              }
+            ],
+            "properties": {
+              "Node name for S&R": "DA3Inference",
+              "cnr_id": "comfy-core",
+              "ver": "0.19.0"
+            },
+            "widgets_values": [
+              504,
+              "lower_bound_resize",
+              "mono"
+            ]
+          },
+          {
+            "id": 94,
+            "type": "LoadDA3Model",
+            "pos": [
+              50,
+              410
+            ],
+            "size": [
+              400,
+              140
+            ],
+            "flags": {},
+            "order": 2,
+            "mode": 0,
+            "inputs": [
+              {
+                "localized_name": "model_name",
+                "name": "model_name",
+                "type": "COMBO",
+                "widget": {
+                  "name": "model_name"
+                },
+                "link": 129
+              },
+              {
+                "localized_name": "weight_dtype",
+                "name": "weight_dtype",
+                "type": "COMBO",
+                "widget": {
+                  "name": "weight_dtype"
+                },
+                "link": null
+              }
+            ],
+            "outputs": [
+              {
+                "localized_name": "DA3_MODEL",
+                "name": "DA3_MODEL",
+                "type": "DA3_MODEL",
+                "links": [
+                  107
+                ]
+              }
+            ],
+            "properties": {
+              "Node name for S&R": "LoadDA3Model",
+              "cnr_id": "comfy-core",
+              "ver": "0.24.0",
+              "models": [
+                {
+                  "name": "depth_anything_3_mono_large.safetensors",
+                  "url": "https://huggingface.co/Comfy-Org/Depth-Anything-3/resolve/main/geometry_estimation/depth_anything_3_mono_large.safetensors",
+                  "directory": "geometry_estimation"
+                }
+              ]
+            },
+            "widgets_values": [
+              "depth_anything_3_mono_large.safetensors",
+              "default"
+            ]
+          },
+          {
+            "id": 95,
+            "type": "GetVideoComponents",
+            "pos": [
+              70,
+              -140
+            ],
+            "size": [
+              260,
+              120
+            ],
+            "flags": {},
+            "order": 3,
+            "mode": 0,
+            "inputs": [
+              {
+                "localized_name": "video",
+                "name": "video",
+                "type": "VIDEO",
+                "link": 120
+              }
+            ],
+            "outputs": [
+              {
+                "localized_name": "images",
+                "name": "images",
+                "type": "IMAGE",
+                "links": [
+                  111
+                ]
+              },
+              {
+                "localized_name": "audio",
+                "name": "audio",
+                "type": "AUDIO",
+                "links": [
+                  112
+                ]
+              },
+              {
+                "localized_name": "fps",
+                "name": "fps",
+                "type": "FLOAT",
+                "links": [
+                  113
+                ]
+              },
+              {
+                "localized_name": "bit_depth",
+                "name": "bit_depth",
+                "type": "INT",
+                "links": null
+              }
+            ],
+            "properties": {
+              "Node name for S&R": "GetVideoComponents",
+              "cnr_id": "comfy-core",
+              "ver": "0.24.0"
+            }
+          },
+          {
+            "id": 96,
+            "type": "Video Slice",
+            "pos": [
+              70,
+              -360
+            ],
+            "size": [
+              270,
+              170
+            ],
+            "flags": {},
+            "order": 4,
+            "mode": 0,
+            "inputs": [
+              {
+                "localized_name": "video",
+                "name": "video",
+                "type": "VIDEO",
+                "link": 119
+              },
+              {
+                "localized_name": "start_time",
+                "name": "start_time",
+                "type": "FLOAT",
+                "widget": {
+                  "name": "start_time"
+                },
+                "link": 121
+              },
+              {
+                "localized_name": "duration",
+                "name": "duration",
+                "type": "FLOAT",
+                "widget": {
+                  "name": "duration"
+                },
+                "link": 122
+              },
+              {
+                "localized_name": "strict_duration",
+                "name": "strict_duration",
+                "type": "BOOLEAN",
+                "widget": {
+                  "name": "strict_duration"
+                },
+                "link": null
+              }
+            ],
+            "outputs": [
+              {
+                "localized_name": "VIDEO",
+                "name": "VIDEO",
+                "type": "VIDEO",
+                "links": [
+                  120
+                ]
+              }
+            ],
+            "properties": {
+              "Node name for S&R": "Video Slice",
+              "cnr_id": "comfy-core",
+              "ver": "0.24.0"
+            },
+            "widgets_values": [
+              0,
+              5,
+              false
+            ]
+          }
+        ],
+        "groups": [],
+        "links": [
+          {
+            "id": 12,
+            "origin_id": 93,
+            "origin_slot": 0,
+            "target_id": 92,
+            "target_slot": 0,
+            "type": "DA3_GEOMETRY"
+          },
+          {
+            "id": 7,
+            "origin_id": 92,
+            "origin_slot": 0,
+            "target_id": -20,
+            "target_slot": 0,
+            "type": "IMAGE"
+          },
+          {
+            "id": 107,
+            "origin_id": 94,
+            "origin_slot": 0,
+            "target_id": 93,
+            "target_slot": 0,
+            "type": "DA3_MODEL"
+          },
+          {
+            "id": 111,
+            "origin_id": 95,
+            "origin_slot": 0,
+            "target_id": 93,
+            "target_slot": 1,
+            "type": "IMAGE"
+          },
+          {
+            "id": 112,
+            "origin_id": 95,
+            "origin_slot": 1,
+            "target_id": -20,
+            "target_slot": 1,
+            "type": "AUDIO"
+          },
+          {
+            "id": 113,
+            "origin_id": 95,
+            "origin_slot": 2,
+            "target_id": -20,
+            "target_slot": 2,
+            "type": "FLOAT"
+          },
+          {
+            "id": 119,
+            "origin_id": -10,
+            "origin_slot": 0,
+            "target_id": 96,
+            "target_slot": 0,
+            "type": "VIDEO"
+          },
+          {
+            "id": 120,
+            "origin_id": 96,
+            "origin_slot": 0,
+            "target_id": 95,
+            "target_slot": 0,
+            "type": "VIDEO"
+          },
+          {
+            "id": 121,
+            "origin_id": -10,
+            "origin_slot": 1,
+            "target_id": 96,
+            "target_slot": 1,
+            "type": "FLOAT"
+          },
+          {
+            "id": 122,
+            "origin_id": -10,
+            "origin_slot": 2,
+            "target_id": 96,
+            "target_slot": 2,
+            "type": "FLOAT"
+          },
+          {
+            "id": 124,
+            "origin_id": -10,
+            "origin_slot": 3,
+            "target_id": 93,
+            "target_slot": 2,
+            "type": "INT"
+          },
+          {
+            "id": 125,
+            "origin_id": -10,
+            "origin_slot": 4,
+            "target_id": 93,
+            "target_slot": 3,
+            "type": "COMBO"
+          },
+          {
+            "id": 126,
+            "origin_id": -10,
+            "origin_slot": 5,
+            "target_id": 92,
+            "target_slot": 1,
+            "type": "COMFY_DYNAMICCOMBO_V3"
+          },
+          {
+            "id": 127,
+            "origin_id": -10,
+            "origin_slot": 6,
+            "target_id": 92,
+            "target_slot": 2,
+            "type": "COMBO"
+          },
+          {
+            "id": 128,
+            "origin_id": -10,
+            "origin_slot": 7,
+            "target_id": 92,
+            "target_slot": 3,
+            "type": "BOOLEAN"
+          },
+          {
+            "id": 129,
+            "origin_id": -10,
+            "origin_slot": 8,
+            "target_id": 94,
+            "target_slot": 0,
+            "type": "COMBO"
+          }
+        ],
+        "extra": {},
+        "category": "Conditioning & Preprocessors/Depth",
+        "description": "This subgraph processes a video input through Depth Anything 3 to produce temporally consistent depth maps for each frame, outputting a depth video. It is ideal for video content requiring spatial geometry estimation, such as 3D reconstruction, SLAM, or novel view synthesis from moving cameras. The model uses a plain transformer backbone trained with a depth-ray representation, supporting any number of views without requiring known camera poses."
+      }
+    ]
+  },
+  "extra": {
+    "BlueprintDescription": "This subgraph processes a video input through Depth Anything 3 to produce temporally consistent depth maps for each frame, outputting a depth video. It is ideal for video content requiring spatial geometry estimation, such as 3D reconstruction, SLAM, or novel view synthesis from moving cameras. The model uses a plain transformer backbone trained with a depth-ray representation, supporting any number of views without requiring known camera poses."
+  }
+}

comfy/background_removal/birefnet.py

@@ -105,7 +105,7 @@ class WindowAttention(nn.Module):
 
         relative_position_bias = self.relative_position_bias_table[self.relative_position_index.long().view(-1)].view(
             self.window_size[0] * self.window_size[1], self.window_size[0] * self.window_size[1], -1)  # Wh*Ww,Wh*Ww,nH
-        relative_position_bias = relative_position_bias.permute(2, 0, 1).contiguous()  # nH, Wh*Ww, Wh*Ww
+        relative_position_bias = comfy.ops.cast_to_input(relative_position_bias.permute(2, 0, 1).contiguous(), attn)  # nH, Wh*Ww, Wh*Ww
         attn = attn + relative_position_bias.unsqueeze(0)
 
         if mask is not None:

comfy/cli_args.py

@@ -115,6 +115,7 @@ cache_group.add_argument("--cache-ram", nargs='*', type=float, default=[], metav
 cache_group.add_argument("--cache-classic", action="store_true", help="Use the old style (aggressive) caching.")
 cache_group.add_argument("--cache-lru", type=int, default=0, help="Use LRU caching with a maximum of N node results cached. May use more RAM/VRAM.")
 cache_group.add_argument("--cache-none", action="store_true", help="Reduced RAM/VRAM usage at the expense of executing every node for each run.")
+cache_group.add_argument("--high-ram", action="store_true", help="Can improve performance slightly on high RAM or on systems where pagefile use is preferred over model loading.")
 
 attn_group = parser.add_mutually_exclusive_group()
 attn_group.add_argument("--use-split-cross-attention", action="store_true", help="Use the split cross attention optimization. Ignored when xformers is used.")
@@ -133,7 +134,7 @@ upcast.add_argument("--dont-upcast-attention", action="store_true", help="Disabl
 parser.add_argument("--enable-manager", action="store_true", help="Enable the ComfyUI-Manager feature.")
 manager_group = parser.add_mutually_exclusive_group()
 manager_group.add_argument("--disable-manager-ui", action="store_true", help="Disables only the ComfyUI-Manager UI and endpoints. Scheduled installations and similar background tasks will still operate.")
-manager_group.add_argument("--enable-manager-legacy-ui", action="store_true", help="Enables the legacy UI of ComfyUI-Manager")
+manager_group.add_argument("--enable-manager-legacy-ui", action="store_true", help="Enables the legacy UI of ComfyUI-Manager. Implies --enable-manager.")
 
 
 vram_group = parser.add_mutually_exclusive_group()
@@ -144,11 +145,13 @@ vram_group.add_argument("--novram", action="store_true", help="When lowvram isn'
 vram_group.add_argument("--cpu", action="store_true", help="To use the CPU for everything (slow).")
 
 parser.add_argument("--reserve-vram", type=float, default=None, help="Set the amount of vram in GB you want to reserve for use by your OS/other software. By default some amount is reserved depending on your OS.")
+parser.add_argument("--vram-headroom", type=float, default=0, help="Set the amount of vram in GB for DynamicVRAM to maintain as extra headroom above default. ComfyUI will try and keep this much VRAM completely free and unused, even counting VRAM from other apps.")
 
 parser.add_argument("--async-offload", nargs='?', const=2, type=int, default=None, metavar="NUM_STREAMS", help="Use async weight offloading. An optional argument controls the amount of offload streams. Default is 2. Enabled by default on Nvidia.")
 parser.add_argument("--disable-async-offload", action="store_true", help="Disable async weight offloading.")
 parser.add_argument("--disable-dynamic-vram", action="store_true", help="Disable dynamic VRAM and use estimate based model loading.")
 parser.add_argument("--enable-dynamic-vram", action="store_true", help="Enable dynamic VRAM on systems where it's not enabled by default.")
+parser.add_argument("--fast-disk", action="store_true", help="Prefer disk-backed dynamic loading and offload over unpinned RAM. Can be faster for users with fast NVME disks.")
 
 parser.add_argument("--force-non-blocking", action="store_true", help="Force ComfyUI to use non-blocking operations for all applicable tensors. This may improve performance on some non-Nvidia systems but can cause issues with some workflows.")
 
@@ -165,6 +168,8 @@ class PerformanceFeature(enum.Enum):
 
 parser.add_argument("--fast", nargs="*", type=PerformanceFeature, help="Enable some untested and potentially quality deteriorating optimizations. This is used to test new features so using it might crash your comfyui. --fast with no arguments enables everything. You can pass a list specific optimizations if you only want to enable specific ones. Current valid optimizations: {}".format(" ".join(map(lambda c: c.value, PerformanceFeature))))
 
+parser.add_argument("--debug-hang", action="store_true", help="Enable stack trace dumps on Ctrl-C for debugging hangs.")
+
 parser.add_argument("--disable-pinned-memory", action="store_true", help="Disable pinned memory use.")
 
 parser.add_argument("--mmap-torch-files", action="store_true", help="Use mmap when loading ckpt/pt files.")
@@ -246,6 +251,9 @@ else:
 if args.cache_ram is not None and len(args.cache_ram) > 2:
     parser.error("--cache-ram accepts at most two values: active GB and inactive GB")
 
+if args.high_ram:
+    args.cache_classic = True
+
 if args.windows_standalone_build:
     args.auto_launch = True
 
@@ -255,6 +263,10 @@ if args.disable_auto_launch:
 if args.force_fp16:
     args.fp16_unet = True
 
+# '--enable-manager-legacy-ui' is meaningless unless the manager is enabled, so imply '--enable-manager'.
+if args.enable_manager_legacy_ui:
+    args.enable_manager = True
+
 
 # '--fast' is not provided, use an empty set
 if args.fast is None:

comfy/clip_vision.py

@@ -9,6 +9,7 @@ import comfy.model_management
 import comfy.utils
 import comfy.clip_model
 import comfy.image_encoders.dino2
+import comfy.image_encoders.dino3
 
 class Output:
     def __getitem__(self, key):
@@ -23,12 +24,16 @@ IMAGE_ENCODERS = {
     "siglip_vision_model": comfy.clip_model.CLIPVisionModelProjection,
     "siglip2_vision_model": comfy.clip_model.CLIPVisionModelProjection,
     "dinov2": comfy.image_encoders.dino2.Dinov2Model,
+    "dinov3": comfy.image_encoders.dino3.DINOv3ViTModel,
 }
 
 class ClipVisionModel():
     def __init__(self, json_config):
-        with open(json_config) as f:
-            config = json.load(f)
+        if isinstance(json_config, dict):
+            config = json_config
+        else:
+            with open(json_config) as f:
+                config = json.load(f)
 
         self.image_size = config.get("image_size", 224)
         self.image_mean = config.get("image_mean", [0.48145466, 0.4578275, 0.40821073])
@@ -134,6 +139,8 @@ def load_clipvision_from_sd(sd, prefix="", convert_keys=False):
         json_config = os.path.join(os.path.join(os.path.dirname(os.path.realpath(__file__)), "image_encoders"), "dino2_giant.json")
     elif 'encoder.layer.23.layer_scale2.lambda1' in sd:
         json_config = os.path.join(os.path.join(os.path.dirname(os.path.realpath(__file__)), "image_encoders"), "dino2_large.json")
+    elif 'layer.0.mlp.gate_proj.weight' in sd and 'layer.31.norm1.weight' in sd: # Dinov3 ViT-H/16+ (SwiGLU gated MLP, 32 layers)
+        json_config = comfy.image_encoders.dino3.DINOV3_VITH_CONFIG
     else:
         return None
 

comfy/context_windows.py

@@ -8,6 +8,8 @@ from abc import ABC, abstractmethod
 import logging
 import comfy.model_management
 import comfy.patcher_extension
+import comfy.utils
+import comfy.conds
 if TYPE_CHECKING:
     from comfy.model_base import BaseModel
     from comfy.model_patcher import ModelPatcher
@@ -51,12 +53,18 @@ class ContextHandlerABC(ABC):
 
 
 class IndexListContextWindow(ContextWindowABC):
-    def __init__(self, index_list: list[int], dim: int=0, total_frames: int=0):
+    def __init__(self, index_list: list[int], dim: int=0, total_frames: int=0, modality_windows: dict=None, context_overlap: int=0):
         self.index_list = index_list
         self.context_length = len(index_list)
+        self.context_overlap = context_overlap
         self.dim = dim
         self.total_frames = total_frames
         self.center_ratio = (min(index_list) + max(index_list)) / (2 * total_frames)
+        self.modality_windows = modality_windows  # dict of {mod_idx: IndexListContextWindow}
+        self.guide_frames_indices: list[int] = []
+        self.guide_overlap_info: list[tuple[int, int]] = []
+        self.guide_kf_local_positions: list[int] = []
+        self.guide_downscale_factors: list[int] = []
 
     def get_tensor(self, full: torch.Tensor, device=None, dim=None, retain_index_list=[]) -> torch.Tensor:
         if dim is None:
@@ -85,6 +93,11 @@ class IndexListContextWindow(ContextWindowABC):
         region_idx = int(self.center_ratio * num_regions)
         return min(max(region_idx, 0), num_regions - 1)
 
+    def get_window_for_modality(self, modality_idx: int) -> 'IndexListContextWindow':
+        if modality_idx == 0:
+            return self
+        return self.modality_windows[modality_idx]
+
 
 class IndexListCallbacks:
     EVALUATE_CONTEXT_WINDOWS = "evaluate_context_windows"
@@ -148,6 +161,172 @@ def slice_cond(cond_value, window: IndexListContextWindow, x_in: torch.Tensor, d
     return cond_value._copy_with(sliced)
 
 
+def compute_guide_overlap(guide_entries: list[dict], keyframe_idxs: torch.Tensor, temporal_downscale_ratio: int, window_index_list: list[int]):
+    """Compute which concatenated guide frames overlap with a context window.
+
+    Each guide's latent-space start is derived from its first token's pixel-t-start
+    in keyframe_idxs (shape (B, [t,h,w], num_tokens, [start, end])), divided by the
+    model's temporal_downscale_ratio.
+
+    Args:
+        guide_entries: list of guide_attention_entry dicts
+        keyframe_idxs: per-token pixel coords cond tensor for the modality
+        temporal_downscale_ratio: model's pixel-to-latent temporal compression ratio
+        window_index_list: the window's frame indices into the video portion
+
+    Returns:
+        suffix_indices: indices into the guide_frames tensor for frame selection
+        overlap_info: list of (entry_idx, overlap_count) for guide_attention_entries adjustment
+        kf_local_positions: window-local frame positions for keyframe_idxs regeneration
+        total_overlap: total number of overlapping guide frames
+    """
+    window_set = set(window_index_list)
+    window_list = list(window_index_list)
+    suffix_indices = []
+    overlap_info = []
+    kf_local_positions = []
+    suffix_base = 0
+    token_offset = 0
+
+    for entry_idx, entry in enumerate(guide_entries):
+        first_t_pixel = int(keyframe_idxs[0, 0, token_offset, 0].item())
+        latent_start = (first_t_pixel + temporal_downscale_ratio - 1) // temporal_downscale_ratio
+        guide_len = entry["latent_shape"][0]
+        entry_overlap = 0
+
+        for local_offset in range(guide_len):
+            video_pos = latent_start + local_offset
+            if video_pos in window_set:
+                suffix_indices.append(suffix_base + local_offset)
+                kf_local_positions.append(window_list.index(video_pos))
+                entry_overlap += 1
+
+        if entry_overlap > 0:
+            overlap_info.append((entry_idx, entry_overlap))
+        suffix_base += guide_len
+        token_offset += entry["pre_filter_count"]
+
+    return suffix_indices, overlap_info, kf_local_positions, len(suffix_indices)
+
+
+@dataclass
+class WindowingState:
+    """Per-modality context windowing state for each step,
+    built using IndexListContextHandler._build_window_state().
+    For non-multimodal models the lists are length 1
+    """
+    latents: list[torch.Tensor]                  # per-modality working latents (guide frames stripped)
+    guide_latents: list[torch.Tensor | None]     # per-modality guide frames stripped from latents
+    guide_entries: list[list[dict] | None]       # per-modality guide_attention_entry metadata
+    keyframe_idxs: list[torch.Tensor | None]     # per-modality keyframe_idxs tensor for guide latent_start derivation
+    latent_shapes: list | None                   # original packed shapes for unpack/pack (None if not multimodal)
+    dim: int = 0                                 # primary modality temporal dim for context windowing
+    is_multimodal: bool = False
+    temporal_downscale_ratio: int = 1            # model's pixel-to-latent temporal compression ratio
+
+    def prepare_window(self, window: IndexListContextWindow, model) -> IndexListContextWindow:
+        """Reformat window for multimodal contexts by deriving per-modality index lists.
+        Non-multimodal contexts return the input window unchanged.
+        """
+        if not self.is_multimodal:
+            return window
+
+        x = self.latents[0]
+        primary_total = self.latent_shapes[0][self.dim]
+        primary_overlap = window.context_overlap
+        map_shapes = self.latent_shapes
+        if x.size(self.dim) != primary_total:
+            map_shapes = list(self.latent_shapes)
+            video_shape = list(self.latent_shapes[0])
+            video_shape[self.dim] = x.size(self.dim)
+            map_shapes[0] = torch.Size(video_shape)
+        try:
+            per_modality_indices = model.map_context_window_to_modalities(
+                window.index_list, map_shapes, self.dim)
+        except AttributeError:
+            raise NotImplementedError(
+                f"{type(model).__name__} must implement map_context_window_to_modalities for multimodal context windows.")
+        modality_windows = {}
+        for mod_idx in range(1, len(self.latents)):
+            modality_total_frames = self.latents[mod_idx].shape[self.dim]
+            ratio = modality_total_frames / primary_total if primary_total > 0 else 1
+            modality_overlap = max(round(primary_overlap * ratio), 0)
+            modality_windows[mod_idx] = IndexListContextWindow(
+                per_modality_indices[mod_idx], dim=self.dim,
+                total_frames=modality_total_frames,
+                context_overlap=modality_overlap)
+        return IndexListContextWindow(
+            window.index_list, dim=self.dim, total_frames=x.shape[self.dim],
+            modality_windows=modality_windows, context_overlap=primary_overlap)
+
+    def slice_for_window(self, window: IndexListContextWindow, retain_index_list: list[int], device=None) -> tuple[list[torch.Tensor], list[int]]:
+        """Slice latents for a context window, injecting guide frames where applicable.
+        For multimodal contexts, uses the modality-specific windows derived in prepare_window().
+        """
+        sliced = []
+        guide_frame_counts = []
+        for idx in range(len(self.latents)):
+            modality_window = window.get_window_for_modality(idx)
+            retain = retain_index_list if idx == 0 else []
+            s = modality_window.get_tensor(self.latents[idx], device, retain_index_list=retain)
+            if self.guide_entries[idx] is not None:
+                s, ng = self._inject_guide_frames(s, modality_window, modality_idx=idx)
+            else:
+                ng = 0
+            sliced.append(s)
+            guide_frame_counts.append(ng)
+        return sliced, guide_frame_counts
+
+    def strip_guide_frames(self, out_per_modality: list[list[torch.Tensor]], guide_frame_counts: list[int], window: IndexListContextWindow):
+        """Strip injected guide frames from per-cond, per-modality outputs in place."""
+        for idx in range(len(self.latents)):
+            if guide_frame_counts[idx] > 0:
+                window_len = len(window.get_window_for_modality(idx).index_list)
+                for ci in range(len(out_per_modality)):
+                    out_per_modality[ci][idx] = out_per_modality[ci][idx].narrow(self.dim, 0, window_len)
+
+    def _inject_guide_frames(self, latent_slice: torch.Tensor, window: IndexListContextWindow, modality_idx: int = 0) -> tuple[torch.Tensor, int]:
+        guide_entries = self.guide_entries[modality_idx]
+        guide_frames = self.guide_latents[modality_idx]
+        keyframe_idxs = self.keyframe_idxs[modality_idx]
+        suffix_idx, overlap_info, kf_local_pos, guide_frame_count = compute_guide_overlap(
+            guide_entries, keyframe_idxs, self.temporal_downscale_ratio, window.index_list)
+        # Shift keyframe positions to account for causal_window_fix anchor occupying sub-pos 0.
+        anchor_idx = getattr(window, 'causal_anchor_index', None)
+        if anchor_idx is not None and anchor_idx >= 0:
+            kf_local_pos = [p + 1 for p in kf_local_pos]
+        window.guide_frames_indices = suffix_idx
+        window.guide_overlap_info = overlap_info
+        window.guide_kf_local_positions = kf_local_pos
+
+        # Derive per-overlap-entry latent_downscale_factor from guide entry latent_shape vs guide frame spatial dims.
+        # guide_frames has full (post-dilation) spatial dims; entry["latent_shape"] has pre-dilation dims.
+        guide_downscale_factors = []
+        if guide_frame_count > 0:
+            full_H = guide_frames.shape[3]
+            for entry_idx, _ in overlap_info:
+                entry_H = guide_entries[entry_idx]["latent_shape"][1]
+                guide_downscale_factors.append(full_H // entry_H)
+        window.guide_downscale_factors = guide_downscale_factors
+
+        if guide_frame_count > 0:
+            idx = tuple([slice(None)] * self.dim + [suffix_idx])
+            return torch.cat([latent_slice, guide_frames[idx]], dim=self.dim), guide_frame_count
+        return latent_slice, 0
+
+    def patch_latent_shapes(self, sub_conds, new_shapes):
+        if not self.is_multimodal:
+            return
+
+        for cond_list in sub_conds:
+            if cond_list is None:
+                continue
+            for cond_dict in cond_list:
+                model_conds = cond_dict.get('model_conds', {})
+                if 'latent_shapes' in model_conds:
+                    model_conds['latent_shapes'] = comfy.conds.CONDConstant(new_shapes)
+
+
 @dataclass
 class ContextSchedule:
     name: str
@@ -162,7 +341,7 @@ ContextResults = collections.namedtuple("ContextResults", ['window_idx', 'sub_co
 class IndexListContextHandler(ContextHandlerABC):
     def __init__(self, context_schedule: ContextSchedule, fuse_method: ContextFuseMethod, context_length: int=1, context_overlap: int=0, context_stride: int=1,
                  closed_loop: bool=False, dim:int=0, freenoise: bool=False, cond_retain_index_list: list[int]=[], split_conds_to_windows: bool=False,
-                 causal_window_fix: bool=True):
+                 latent_retain_index_list: list[int]=[], causal_window_fix: bool=True):
         self.context_schedule = context_schedule
         self.fuse_method = fuse_method
         self.context_length = context_length
@@ -174,17 +353,118 @@ class IndexListContextHandler(ContextHandlerABC):
         self.freenoise = freenoise
         self.cond_retain_index_list = [int(x.strip()) for x in cond_retain_index_list.split(",")] if cond_retain_index_list else []
         self.split_conds_to_windows = split_conds_to_windows
+        self.latent_retain_index_list = [int(x.strip()) for x in latent_retain_index_list.split(",")] if latent_retain_index_list else []
         self.causal_window_fix = causal_window_fix
 
         self.callbacks = {}
 
+    @staticmethod
+    def _get_latent_shapes(conds):
+        for cond_list in conds:
+            if cond_list is None:
+                continue
+            for cond_dict in cond_list:
+                model_conds = cond_dict.get('model_conds', {})
+                if 'latent_shapes' in model_conds:
+                    return model_conds['latent_shapes'].cond
+        return None
+
+    @staticmethod
+    def _get_guide_entries(conds):
+        for cond_list in conds:
+            if cond_list is None:
+                continue
+            for cond_dict in cond_list:
+                model_conds = cond_dict.get('model_conds', {})
+                entries = model_conds.get('guide_attention_entries')
+                if entries is not None and hasattr(entries, 'cond') and entries.cond:
+                    return entries.cond
+        return None
+
+    @staticmethod
+    def _get_keyframe_idxs(conds):
+        for cond_list in conds:
+            if cond_list is None:
+                continue
+            for cond_dict in cond_list:
+                model_conds = cond_dict.get('model_conds', {})
+                kf = model_conds.get('keyframe_idxs')
+                if kf is not None and hasattr(kf, 'cond') and kf.cond is not None:
+                    return kf.cond
+        return None
+
+    def _apply_freenoise(self, noise: torch.Tensor, conds: list[list[dict]], seed: int) -> torch.Tensor:
+        """Apply FreeNoise shuffling, scaling context length/overlap per-modality by frame ratio.
+        If guide frames are present on the primary modality, only the video portion is shuffled.
+        """
+        guide_entries = self._get_guide_entries(conds)
+        guide_count = sum(e["latent_shape"][0] for e in guide_entries) if guide_entries else 0
+
+        latent_shapes = self._get_latent_shapes(conds)
+        if latent_shapes is not None and len(latent_shapes) > 1:
+            modalities = comfy.utils.unpack_latents(noise, latent_shapes)
+            primary_total = latent_shapes[0][self.dim]
+            primary_video_count = modalities[0].size(self.dim) - guide_count
+            apply_freenoise(modalities[0].narrow(self.dim, 0, primary_video_count), self.dim, self.context_length, self.context_overlap, seed)
+            for i in range(1, len(modalities)):
+                mod_total = latent_shapes[i][self.dim]
+                ratio = mod_total / primary_total if primary_total > 0 else 1
+                mod_ctx_len = max(round(self.context_length * ratio), 1)
+                mod_ctx_overlap = max(round(self.context_overlap * ratio), 0)
+                modalities[i] = apply_freenoise(modalities[i], self.dim, mod_ctx_len, mod_ctx_overlap, seed)
+            noise, _ = comfy.utils.pack_latents(modalities)
+            return noise
+        video_count = noise.size(self.dim) - guide_count
+        apply_freenoise(noise.narrow(self.dim, 0, video_count), self.dim, self.context_length, self.context_overlap, seed)
+        return noise
+
+    def _build_window_state(self, x_in: torch.Tensor, conds: list[list[dict]], model: BaseModel) -> WindowingState:
+        """Build windowing state for the current step, including unpacking latents and extracting guide frame info from conds."""
+        latent_shapes = self._get_latent_shapes(conds)
+        is_multimodal = latent_shapes is not None and len(latent_shapes) > 1
+        unpacked_latents = comfy.utils.unpack_latents(x_in, latent_shapes) if is_multimodal else [x_in]
+
+        unpacked_latents_list = list(unpacked_latents)
+        guide_latents_list = [None] * len(unpacked_latents)
+        guide_entries_list = [None] * len(unpacked_latents)
+        keyframe_idxs_list = [None] * len(unpacked_latents)
+
+        extracted_guide_entries = self._get_guide_entries(conds)
+        extracted_keyframe_idxs = self._get_keyframe_idxs(conds)
+
+        # Strip guide frames (only from first modality for now)
+        if extracted_guide_entries is not None:
+            guide_count = sum(e["latent_shape"][0] for e in extracted_guide_entries)
+            if guide_count > 0:
+                x = unpacked_latents[0]
+                latent_count = x.size(self.dim) - guide_count
+                unpacked_latents_list[0] = x.narrow(self.dim, 0, latent_count)
+                guide_latents_list[0] = x.narrow(self.dim, latent_count, guide_count)
+                guide_entries_list[0] = extracted_guide_entries
+                keyframe_idxs_list[0] = extracted_keyframe_idxs
+
+
+        return WindowingState(
+            latents=unpacked_latents_list,
+            guide_latents=guide_latents_list,
+            guide_entries=guide_entries_list,
+            keyframe_idxs=keyframe_idxs_list,
+            latent_shapes=latent_shapes,
+            dim=self.dim,
+            is_multimodal=is_multimodal,
+            temporal_downscale_ratio=model.latent_format.temporal_downscale_ratio)
+
     def should_use_context(self, model: BaseModel, conds: list[list[dict]], x_in: torch.Tensor, timestep: torch.Tensor, model_options: dict[str]) -> bool:
-        # for now, assume first dim is batch - should have stored on BaseModel in actual implementation
-        if x_in.size(self.dim) > self.context_length:
-            logging.info(f"Using context windows {self.context_length} with overlap {self.context_overlap} for {x_in.size(self.dim)} frames.")
+        window_state = self._build_window_state(x_in, conds, model) # build window_state to check frame counts, will be built again in execute
+        total_frame_count = window_state.latents[0].size(self.dim)
+        if total_frame_count > self.context_length:
+            logging.info(f"\nUsing context windows: Context length {self.context_length} with overlap {self.context_overlap} for {total_frame_count} frames.")
             if self.cond_retain_index_list:
                 logging.info(f"Retaining original cond for indexes: {self.cond_retain_index_list}")
+            if self.latent_retain_index_list:
+                logging.info(f"Retaining original latent for indexes: {self.latent_retain_index_list}")
             return True
+        logging.info(f"\nNot using context windows since context length ({self.context_length}) exceeds input frames ({total_frame_count}).")
         return False
 
     def prepare_control_objects(self, control: ControlBase, device=None) -> ControlBase:
@@ -275,7 +555,9 @@ class IndexListContextHandler(ContextHandlerABC):
         return resized_cond
 
     def set_step(self, timestep: torch.Tensor, model_options: dict[str]):
-        mask = torch.isclose(model_options["transformer_options"]["sample_sigmas"], timestep[0], rtol=0.0001)
+        sample_sigmas = model_options["transformer_options"]["sample_sigmas"]
+        current_timestep = timestep[0].to(sample_sigmas.dtype)
+        mask = torch.isclose(sample_sigmas, current_timestep, rtol=0.0001)
         matches = torch.nonzero(mask)
         if torch.numel(matches) == 0:
             return  # substep from multi-step sampler: keep self._step from the last full step
@@ -284,54 +566,98 @@ class IndexListContextHandler(ContextHandlerABC):
     def get_context_windows(self, model: BaseModel, x_in: torch.Tensor, model_options: dict[str]) -> list[IndexListContextWindow]:
         full_length = x_in.size(self.dim) # TODO: choose dim based on model
         context_windows = self.context_schedule.func(full_length, self, model_options)
-        context_windows = [IndexListContextWindow(window, dim=self.dim, total_frames=full_length) for window in context_windows]
+        context_windows = [IndexListContextWindow(window, dim=self.dim, total_frames=full_length, context_overlap=self.context_overlap) for window in context_windows]
         return context_windows
 
     def execute(self, calc_cond_batch: Callable, model: BaseModel, conds: list[list[dict]], x_in: torch.Tensor, timestep: torch.Tensor, model_options: dict[str]):
         self._model = model
         self.set_step(timestep, model_options)
-        context_windows = self.get_context_windows(model, x_in, model_options)
-        enumerated_context_windows = list(enumerate(context_windows))
 
-        conds_final = [torch.zeros_like(x_in) for _ in conds]
+        window_state = self._build_window_state(x_in, conds, model)
+        num_modalities = len(window_state.latents)
+
+        context_windows = self.get_context_windows(model, window_state.latents[0], model_options)
+        enumerated_context_windows = list(enumerate(context_windows))
+        total_windows = len(enumerated_context_windows)
+
+        # Initialize per-modality accumulators (length 1 for single-modality)
+        accum = [[torch.zeros_like(m) for _ in conds] for m in window_state.latents]
         if self.fuse_method.name == ContextFuseMethods.RELATIVE:
-            counts_final = [torch.ones(get_shape_for_dim(x_in, self.dim), device=x_in.device) for _ in conds]
+            counts = [[torch.ones(get_shape_for_dim(m, self.dim), device=m.device) for _ in conds] for m in window_state.latents]
         else:
-            counts_final = [torch.zeros(get_shape_for_dim(x_in, self.dim), device=x_in.device) for _ in conds]
-        biases_final = [([0.0] * x_in.shape[self.dim]) for _ in conds]
+            counts = [[torch.zeros(get_shape_for_dim(m, self.dim), device=m.device) for _ in conds] for m in window_state.latents]
+        biases = [[([0.0] * m.shape[self.dim]) for _ in conds] for m in window_state.latents]
 
         for callback in comfy.patcher_extension.get_all_callbacks(IndexListCallbacks.EXECUTE_START, self.callbacks):
             callback(self, model, x_in, conds, timestep, model_options)
 
+        # accumulate results from each context window
         for enum_window in enumerated_context_windows:
-            results = self.evaluate_context_windows(calc_cond_batch, model, x_in, conds, timestep, [enum_window], model_options)
+            results = self.evaluate_context_windows(
+                calc_cond_batch, model, x_in, conds, timestep, [enum_window],
+                model_options, window_state=window_state, total_windows=total_windows)
             for result in results:
-                self.combine_context_window_results(x_in, result.sub_conds_out, result.sub_conds, result.window, result.window_idx, len(enumerated_context_windows), timestep,
-                                            conds_final, counts_final, biases_final)
+                # result.sub_conds_out is per-cond, per-modality: list[list[Tensor]]
+                for mod_idx in range(num_modalities):
+                    mod_out = [result.sub_conds_out[ci][mod_idx] for ci in range(len(conds))]
+                    modality_window = result.window.get_window_for_modality(mod_idx)
+                    self.combine_context_window_results(
+                        window_state.latents[mod_idx], mod_out, result.sub_conds, modality_window,
+                        result.window_idx, total_windows, timestep,
+                        accum[mod_idx], counts[mod_idx], biases[mod_idx])
+
+        # fuse accumulated results into final conds
         try:
-            # finalize conds
-            if self.fuse_method.name == ContextFuseMethods.RELATIVE:
-                # relative is already normalized, so return as is
-                del counts_final
-                return conds_final
-            else:
-                # normalize conds via division by context usage counts
-                for i in range(len(conds_final)):
-                    conds_final[i] /= counts_final[i]
-                del counts_final
-                return conds_final
+            result_out = []
+            for ci in range(len(conds)):
+                finalized = []
+                for mod_idx in range(num_modalities):
+                    if self.fuse_method.name != ContextFuseMethods.RELATIVE:
+                        accum[mod_idx][ci] /= counts[mod_idx][ci]
+                    f = accum[mod_idx][ci]
+
+                    # if guide frames were injected, append them to the end of the fused latents for the next step
+                    if window_state.guide_latents[mod_idx] is not None:
+                        f = torch.cat([f, window_state.guide_latents[mod_idx]], dim=self.dim)
+                    finalized.append(f)
+
+                # pack modalities together if needed
+                if window_state.is_multimodal and len(finalized) > 1:
+                    packed, _ = comfy.utils.pack_latents(finalized)
+                else:
+                    packed = finalized[0]
+
+                result_out.append(packed)
+            return result_out
         finally:
             for callback in comfy.patcher_extension.get_all_callbacks(IndexListCallbacks.EXECUTE_CLEANUP, self.callbacks):
                 callback(self, model, x_in, conds, timestep, model_options)
 
-    def evaluate_context_windows(self, calc_cond_batch: Callable, model: BaseModel, x_in: torch.Tensor, conds, timestep: torch.Tensor, enumerated_context_windows: list[tuple[int, IndexListContextWindow]],
-                                model_options, device=None, first_device=None):
+    def evaluate_context_windows(self, calc_cond_batch: Callable, model: BaseModel, x_in: torch.Tensor, conds,
+                                timestep: torch.Tensor, enumerated_context_windows: list[tuple[int, IndexListContextWindow]],
+                                model_options, window_state: WindowingState, total_windows: int = None,
+                                device=None, first_device=None):
+        """Evaluate context windows and return per-cond, per-modality outputs in ContextResults.sub_conds_out
+
+        For each window:
+        1. Builds windows (for each modality if multimodal)
+        2. Slices window for each modality
+        3. Injects concatenated latent guide frames where present
+        4. Packs together if needed and calls model
+        5. Unpacks and strips any guides from outputs
+        """
+        x = window_state.latents[0]
+
         results: list[ContextResults] = []
         for window_idx, window in enumerated_context_windows:
             # allow processing to end between context window executions for faster Cancel
             comfy.model_management.throw_exception_if_processing_interrupted()
 
-            # causal_window_fix: prepend a pre-window frame that will be stripped post-forward
+            # prepare the window accounting for multimodal windows
+            window = window_state.prepare_window(window, model)
+
+            # causal_window_fix: prepend a pre-window frame that will be stripped post-forward.
+            # Set anchor before slice_for_window so the latent slice and downstream cond slices both pick it up.
             anchor_applied = False
             if self.causal_window_fix:
                 anchor_idx = window.index_list[0] - 1
@@ -339,27 +665,46 @@ class IndexListContextHandler(ContextHandlerABC):
                     window.causal_anchor_index = anchor_idx
                     anchor_applied = True
 
+            # slice the window for each modality, injecting guide frames where applicable
+            sliced, guide_frame_counts_per_modality = window_state.slice_for_window(window, self.latent_retain_index_list, device)
+
             for callback in comfy.patcher_extension.get_all_callbacks(IndexListCallbacks.EVALUATE_CONTEXT_WINDOWS, self.callbacks):
                 callback(self, model, x_in, conds, timestep, model_options, window_idx, window, model_options, device, first_device)
 
-            # update exposed params
+            logging.info(f"Context window {window_idx + 1}/{total_windows or len(enumerated_context_windows)}: frames {window.index_list[0]}-{window.index_list[-1]} of {x.shape[self.dim]}"
+                         + (f" (+{guide_frame_counts_per_modality[0]} guide frames)" if guide_frame_counts_per_modality[0] > 0 else "")
+                         )
+
+            # if multimodal, pack modalities together
+            if window_state.is_multimodal and len(sliced) > 1:
+                sub_x, sub_shapes = comfy.utils.pack_latents(sliced)
+            else:
+                sub_x, sub_shapes = sliced[0], [sliced[0].shape]
+
+            # get resized conds for window
             model_options["transformer_options"]["context_window"] = window
-            # get subsections of x, timestep, conds
-            sub_x = window.get_tensor(x_in, device)
-            sub_timestep = window.get_tensor(timestep, device, dim=0)
-            sub_conds = [self.get_resized_cond(cond, x_in, window, device) for cond in conds]
+            sub_timestep = window.get_tensor(timestep, dim=0)
+            sub_conds = [self.get_resized_cond(cond, x, window) for cond in conds]
 
+            # if multimodal, patch latent_shapes in conds for correct unpacking in model
+            window_state.patch_latent_shapes(sub_conds, sub_shapes)
+
+            # call model on window
             sub_conds_out = calc_cond_batch(model, sub_conds, sub_x, sub_timestep, model_options)
-            if device is not None:
-                for i in range(len(sub_conds_out)):
-                    sub_conds_out[i] = sub_conds_out[i].to(x_in.device)
 
-            # strip causal_window_fix anchor if applied
+            # unpack outputs
+            out_per_modality = [comfy.utils.unpack_latents(sub_conds_out[i], sub_shapes) for i in range(len(sub_conds_out))]
+
+            # strip causal_window_fix anchor from primary modality before guide strip so window_len math stays correct
             if anchor_applied:
-                for i in range(len(sub_conds_out)):
-                    sub_conds_out[i] = sub_conds_out[i].narrow(self.dim, 1, sub_conds_out[i].shape[self.dim] - 1)
+                for ci in range(len(out_per_modality)):
+                    t = out_per_modality[ci][0]
+                    out_per_modality[ci][0] = t.narrow(self.dim, 1, t.shape[self.dim] - 1)
 
-            results.append(ContextResults(window_idx, sub_conds_out, sub_conds, window))
+            # strip injected guide frames
+            window_state.strip_guide_frames(out_per_modality, guide_frame_counts_per_modality, window)
+
+            results.append(ContextResults(window_idx, out_per_modality, sub_conds, window))
         return results
 
 
@@ -383,7 +728,7 @@ class IndexListContextHandler(ContextHandlerABC):
                     biases_final[i][idx] = bias_total + bias
         else:
             # add conds and counts based on weights of fuse method
-            weights = get_context_weights(window.context_length, x_in.shape[self.dim], window.index_list, self, sigma=timestep)
+            weights = get_context_weights(window.context_length, x_in.shape[self.dim], window.index_list, self, sigma=timestep, context_overlap=window.context_overlap)
             weights_tensor = match_weights_to_dim(weights, x_in, self.dim, device=x_in.device)
             for i in range(len(sub_conds_out)):
                 window.add_window(conds_final[i], sub_conds_out[i] * weights_tensor)
@@ -393,16 +738,22 @@ class IndexListContextHandler(ContextHandlerABC):
             callback(self, x_in, sub_conds_out, sub_conds, window, window_idx, total_windows, timestep, conds_final, counts_final, biases_final)
 
 
-def _prepare_sampling_wrapper(executor, model, noise_shape: torch.Tensor, *args, **kwargs):
-    # limit noise_shape length to context_length for more accurate vram use estimation
+def _prepare_sampling_wrapper(executor, model, noise_shape: torch.Tensor, conds, *args, **kwargs):
+    # Scale noise_shape to a single context window so VRAM estimation budgets per-window.
     model_options = kwargs.get("model_options", None)
     if model_options is None:
         raise Exception("model_options not found in prepare_sampling_wrapper; this should never happen, something went wrong.")
     handler: IndexListContextHandler = model_options.get("context_handler", None)
     if handler is not None:
         noise_shape = list(noise_shape)
-        noise_shape[handler.dim] = min(noise_shape[handler.dim], handler.context_length)
-    return executor(model, noise_shape, *args, **kwargs)
+        is_packed = len(noise_shape) == 3 and noise_shape[1] == 1
+        if is_packed:
+            # TODO: latent_shapes cond isn't attached yet at this point, so we can't compute a
+            # per-window flat latent here. Skipping the clamp over-estimates but prevents immediate OOM.
+            pass
+        elif handler.dim < len(noise_shape) and noise_shape[handler.dim] > handler.context_length:
+            noise_shape[handler.dim] = min(noise_shape[handler.dim], handler.context_length)
+    return executor(model, noise_shape, conds, *args, **kwargs)
 
 
 def create_prepare_sampling_wrapper(model: ModelPatcher):
@@ -422,11 +773,12 @@ def _sampler_sample_wrapper(executor, guider, sigmas, extra_args, callback, nois
         raise Exception("context_handler not found in sampler_sample_wrapper; this should never happen, something went wrong.")
     if not handler.freenoise:
         return executor(guider, sigmas, extra_args, callback, noise, *args, **kwargs)
-    noise = apply_freenoise(noise, handler.dim, handler.context_length, handler.context_overlap, extra_args["seed"])
+
+    conds = [guider.conds.get('positive', guider.conds.get('negative', []))]
+    noise = handler._apply_freenoise(noise, conds, extra_args["seed"])
 
     return executor(guider, sigmas, extra_args, callback, noise, *args, **kwargs)
 
-
 def create_sampler_sample_wrapper(model: ModelPatcher):
     model.add_wrapper_with_key(
         comfy.patcher_extension.WrappersMP.SAMPLER_SAMPLE,
@@ -434,7 +786,6 @@ def create_sampler_sample_wrapper(model: ModelPatcher):
         _sampler_sample_wrapper
     )
 
-
 def match_weights_to_dim(weights: list[float], x_in: torch.Tensor, dim: int, device=None) -> torch.Tensor:
     total_dims = len(x_in.shape)
     weights_tensor = torch.Tensor(weights).to(device=device)
@@ -580,8 +931,9 @@ def get_matching_context_schedule(context_schedule: str) -> ContextSchedule:
     return ContextSchedule(context_schedule, func)
 
 
-def get_context_weights(length: int, full_length: int, idxs: list[int], handler: IndexListContextHandler, sigma: torch.Tensor=None):
-    return handler.fuse_method.func(length, sigma=sigma, handler=handler, full_length=full_length, idxs=idxs)
+def get_context_weights(length: int, full_length: int, idxs: list[int], handler: IndexListContextHandler, sigma: torch.Tensor=None, context_overlap: int=None):
+    context_overlap = handler.context_overlap if context_overlap is None else context_overlap
+    return handler.fuse_method.func(length, sigma=sigma, handler=handler, full_length=full_length, idxs=idxs, context_overlap=context_overlap)
 
 
 def create_weights_flat(length: int, **kwargs) -> list[float]:
@@ -599,18 +951,18 @@ def create_weights_pyramid(length: int, **kwargs) -> list[float]:
         weight_sequence = list(range(1, max_weight, 1)) + [max_weight] + list(range(max_weight - 1, 0, -1))
     return weight_sequence
 
-def create_weights_overlap_linear(length: int, full_length: int, idxs: list[int], handler: IndexListContextHandler, **kwargs):
+def create_weights_overlap_linear(length: int, full_length: int, idxs: list[int], context_overlap: int, **kwargs):
     # based on code in Kijai's WanVideoWrapper: https://github.com/kijai/ComfyUI-WanVideoWrapper/blob/dbb2523b37e4ccdf45127e5ae33e31362f755c8e/nodes.py#L1302
     # only expected overlap is given different weights
     weights_torch = torch.ones((length))
     # blend left-side on all except first window
     if min(idxs) > 0:
-        ramp_up = torch.linspace(1e-37, 1, handler.context_overlap)
-        weights_torch[:handler.context_overlap] = ramp_up
+        ramp_up = torch.linspace(1e-37, 1, context_overlap)
+        weights_torch[:context_overlap] = ramp_up
     # blend right-side on all except last window
     if max(idxs) < full_length-1:
-        ramp_down = torch.linspace(1, 1e-37, handler.context_overlap)
-        weights_torch[-handler.context_overlap:] = ramp_down
+        ramp_down = torch.linspace(1, 1e-37, context_overlap)
+        weights_torch[-context_overlap:] = ramp_down
     return weights_torch
 
 class ContextFuseMethods:

comfy/image_encoders/dino2.py

@@ -1,7 +1,13 @@
 import torch
+import torch.nn.functional as F
+
 from comfy.text_encoders.bert import BertAttention
 import comfy.model_management
 from comfy.ldm.modules.attention import optimized_attention_for_device
+from comfy.ldm.depth_anything_3.reference_view_selector import (
+    select_reference_view, reorder_by_reference, restore_original_order,
+    THRESH_FOR_REF_SELECTION,
+)
 
 
 class Dino2AttentionOutput(torch.nn.Module):
@@ -14,13 +20,41 @@ class Dino2AttentionOutput(torch.nn.Module):
 
 
 class Dino2AttentionBlock(torch.nn.Module):
-    def __init__(self, embed_dim, heads, layer_norm_eps, dtype, device, operations):
+    def __init__(self, embed_dim, heads, layer_norm_eps, dtype, device, operations,
+                 qk_norm=False):
         super().__init__()
+        self.heads = heads
+        self.head_dim = embed_dim // heads
         self.attention = BertAttention(embed_dim, heads, dtype, device, operations)
         self.output = Dino2AttentionOutput(embed_dim, embed_dim, layer_norm_eps, dtype, device, operations)
+        if qk_norm:
+            self.q_norm = operations.LayerNorm(self.head_dim, dtype=dtype, device=device)
+            self.k_norm = operations.LayerNorm(self.head_dim, dtype=dtype, device=device)
+        else:
+            self.q_norm = None
+            self.k_norm = None
 
-    def forward(self, x, mask, optimized_attention):
-        return self.output(self.attention(x, mask, optimized_attention))
+    def forward(self, x, mask, optimized_attention, pos=None, rope=None):
+        # Fast path used by the existing CLIP-vision DINOv2 (no DA3 extensions).
+        if self.q_norm is None and rope is None:
+            return self.output(self.attention(x, mask, optimized_attention))
+
+        # DA3 path: do QKV manually so we can apply per-head QK-norm and 2D RoPE.
+        attn = self.attention
+        B, N, C = x.shape
+        h = self.heads
+        d = self.head_dim
+        q = attn.query(x).view(B, N, h, d).transpose(1, 2)
+        k = attn.key(x).view(B, N, h, d).transpose(1, 2)
+        v = attn.value(x).view(B, N, h, d).transpose(1, 2)
+        if self.q_norm is not None:
+            q = self.q_norm(q)
+            k = self.k_norm(k)
+        if rope is not None and pos is not None:
+            q = rope(q, pos)
+            k = rope(k, pos)
+        out = optimized_attention(q, k, v, h, mask=mask, skip_reshape=True)
+        return self.output(out)
 
 
 class LayerScale(torch.nn.Module):
@@ -64,9 +98,11 @@ class SwiGLUFFN(torch.nn.Module):
 
 
 class Dino2Block(torch.nn.Module):
-    def __init__(self, dim, num_heads, layer_norm_eps, dtype, device, operations, use_swiglu_ffn):
+    def __init__(self, dim, num_heads, layer_norm_eps, dtype, device, operations, use_swiglu_ffn,
+                 qk_norm=False):
         super().__init__()
-        self.attention = Dino2AttentionBlock(dim, num_heads, layer_norm_eps, dtype, device, operations)
+        self.attention = Dino2AttentionBlock(dim, num_heads, layer_norm_eps, dtype, device, operations,
+                                             qk_norm=qk_norm)
         self.layer_scale1 = LayerScale(dim, dtype, device, operations)
         self.layer_scale2 = LayerScale(dim, dtype, device, operations)
         if use_swiglu_ffn:
@@ -76,19 +112,90 @@ class Dino2Block(torch.nn.Module):
         self.norm1 = operations.LayerNorm(dim, eps=layer_norm_eps, dtype=dtype, device=device)
         self.norm2 = operations.LayerNorm(dim, eps=layer_norm_eps, dtype=dtype, device=device)
 
-    def forward(self, x, optimized_attention):
-        x = x + self.layer_scale1(self.attention(self.norm1(x), None, optimized_attention))
+    def forward(self, x, optimized_attention, pos=None, rope=None, attn_mask=None):
+        x = x + self.layer_scale1(self.attention(self.norm1(x), attn_mask, optimized_attention,
+                                                 pos=pos, rope=rope))
         x = x + self.layer_scale2(self.mlp(self.norm2(x)))
         return x
 
 
-class Dino2Encoder(torch.nn.Module):
-    def __init__(self, dim, num_heads, layer_norm_eps, num_layers, dtype, device, operations, use_swiglu_ffn):
+# -----------------------------------------------------------------------------
+# 2D Rotary position embedding (DA3 extension)
+# -----------------------------------------------------------------------------
+
+
+class _PositionGetter:
+    """Cache (h, w) -> flat (y, x) position grid used to feed ``rope``."""
+
+    def __init__(self):
+        self._cache: dict = {}
+
+    def __call__(self, batch_size: int, height: int, width: int, device) -> torch.Tensor:
+        key = (height, width, device)
+        if key not in self._cache:
+            y = torch.arange(height, device=device)
+            x = torch.arange(width, device=device)
+            self._cache[key] = torch.cartesian_prod(y, x)
+        cached = self._cache[key]
+        return cached.view(1, height * width, 2).expand(batch_size, -1, -1).clone()
+
+
+class RotaryPositionEmbedding2D(torch.nn.Module):
+    """2D RoPE used by DA3-Small/Base. No learnable parameters."""
+
+    def __init__(self, frequency: float = 100.0):
         super().__init__()
-        self.layer = torch.nn.ModuleList([Dino2Block(dim, num_heads, layer_norm_eps, dtype, device, operations, use_swiglu_ffn = use_swiglu_ffn)
-                                          for _ in range(num_layers)])
+        self.base_frequency = frequency
+        self._freq_cache: dict = {}
+
+    def _components(self, dim: int, seq_len: int, device, dtype):
+        key = (dim, seq_len, device, dtype)
+        if key not in self._freq_cache:
+            exp = torch.arange(0, dim, 2, device=device).float() / dim
+            inv_freq = 1.0 / (self.base_frequency ** exp)
+            pos = torch.arange(seq_len, device=device, dtype=inv_freq.dtype)
+            ang = torch.einsum("i,j->ij", pos, inv_freq)
+            ang = ang.to(dtype)
+            ang = torch.cat((ang, ang), dim=-1)
+            self._freq_cache[key] = (ang.cos().to(dtype), ang.sin().to(dtype))
+        return self._freq_cache[key]
+
+    @staticmethod
+    def _rotate(x: torch.Tensor) -> torch.Tensor:
+        d = x.shape[-1]
+        x1, x2 = x[..., : d // 2], x[..., d // 2:]
+        return torch.cat((-x2, x1), dim=-1)
+
+    def _apply_1d(self, tokens, positions, cos_c, sin_c):
+        cos = F.embedding(positions, cos_c)[:, None, :, :]
+        sin = F.embedding(positions, sin_c)[:, None, :, :]
+        return (tokens * cos) + (self._rotate(tokens) * sin)
+
+    def forward(self, tokens: torch.Tensor, positions: torch.Tensor) -> torch.Tensor:
+        feature_dim = tokens.size(-1) // 2
+        max_pos = int(positions.max()) + 1
+        cos_c, sin_c = self._components(feature_dim, max_pos, tokens.device, tokens.dtype)
+        v, h = tokens.chunk(2, dim=-1)
+        v = self._apply_1d(v, positions[..., 0], cos_c, sin_c)
+        h = self._apply_1d(h, positions[..., 1], cos_c, sin_c)
+        return torch.cat((v, h), dim=-1)
+
+
+class Dino2Encoder(torch.nn.Module):
+    def __init__(self, dim, num_heads, layer_norm_eps, num_layers, dtype, device, operations, use_swiglu_ffn,
+                 qknorm_start: int = -1):
+        super().__init__()
+        self.layer = torch.nn.ModuleList([
+            Dino2Block(
+                dim, num_heads, layer_norm_eps, dtype, device, operations,
+                use_swiglu_ffn=use_swiglu_ffn,
+                qk_norm=(qknorm_start != -1 and i >= qknorm_start),
+            )
+            for i in range(num_layers)
+        ])
 
     def forward(self, x, intermediate_output=None):
+        # Backward-compat path used by ``ClipVisionModel`` (no DA3 extensions).
         optimized_attention = optimized_attention_for_device(x.device, False, small_input=True)
 
         if intermediate_output is not None:
@@ -122,16 +229,27 @@ class Dino2PatchEmbeddings(torch.nn.Module):
 
 
 class Dino2Embeddings(torch.nn.Module):
-    def __init__(self, dim, dtype, device, operations):
+    def __init__(self, dim, dtype, device, operations,
+                 patch_size: int = 14, image_size: int = 518,
+                 use_mask_token: bool = True,
+                 num_camera_tokens: int = 0):
         super().__init__()
-        patch_size = 14
-        image_size = 518
         self.patch_size = patch_size
+        self.image_size = image_size
 
         self.patch_embeddings = Dino2PatchEmbeddings(dim, patch_size=patch_size, image_size=image_size, dtype=dtype, device=device, operations=operations)
         self.position_embeddings = torch.nn.Parameter(torch.empty(1, (image_size // patch_size) ** 2 + 1, dim, dtype=dtype, device=device))
         self.cls_token = torch.nn.Parameter(torch.empty(1, 1, dim, dtype=dtype, device=device)) # mask_token is a pre-training param, kept only so strict loading accepts the key.
-        self.mask_token = torch.nn.Parameter(torch.empty(1, dim, dtype=dtype, device=device))
+        if use_mask_token:
+            self.mask_token = torch.nn.Parameter(torch.empty(1, dim, dtype=dtype, device=device))
+        else:
+            self.mask_token = None
+        if num_camera_tokens > 0:
+            # DA3 stores (ref_token, src_token) pairs that get injected at the
+            # alt-attn boundary; see ``Dinov2Model._inject_camera_token``.
+            self.camera_token = torch.nn.Parameter(torch.empty(1, num_camera_tokens, dim, dtype=dtype, device=device))
+        else:
+            self.camera_token = None
 
     def interpolate_pos_encoding(self, x, h_pixels, w_pixels):
         pos_embed = comfy.model_management.cast_to_device(self.position_embeddings, x.device, torch.float32)
@@ -140,12 +258,22 @@ class Dino2Embeddings(torch.nn.Module):
         patch_pos = pos_embed[:, 1:]
         N = patch_pos.shape[1]
         M = int(N ** 0.5)
+        assert N == M * M, f"DINOv2 position grid must be square, got N={N} patches (sqrt={M})"
         h0 = h_pixels // self.patch_size
         w0 = w_pixels // self.patch_size
-        scale_factor = ((h0 + 0.1) / M, (w0 + 0.1) / M)  # +0.1 matches upstream DINOv2's FP-rounding workaround so the interpolate output size lands on (h0, w0).
+        # +0.1 matches upstream DINOv2's FP-rounding workaround so the interpolate output size lands on (h0, w0).
+        # scale_factor is (height_scale, width_scale) -- height MUST come first;
+        # swapping these only happens to work for square inputs and breaks
+        # non-square paths like DA3-Small / DA3-Base multi-view.
+        scale_factor = ((h0 + 0.1) / M, (w0 + 0.1) / M)
 
         patch_pos = patch_pos.reshape(1, M, M, -1).permute(0, 3, 1, 2)
         patch_pos = torch.nn.functional.interpolate(patch_pos, scale_factor=scale_factor, mode="bicubic", antialias=False)
+        assert (h0, w0) == patch_pos.shape[-2:], (
+            f"Interpolated pos-embed grid {tuple(patch_pos.shape[-2:])} does not match "
+            f"target patch grid ({h0}, {w0}) for input {h_pixels}x{w_pixels} (patch_size={self.patch_size}); "
+            f"check scale_factor axis order and +0.1 rounding workaround"
+        )
         patch_pos = patch_pos.permute(0, 2, 3, 1).flatten(1, 2)
         return torch.cat((class_pos, patch_pos), dim=1).to(x.dtype)
 
@@ -168,12 +296,51 @@ class Dinov2Model(torch.nn.Module):
         heads = config_dict["num_attention_heads"]
         layer_norm_eps = config_dict["layer_norm_eps"]
         use_swiglu_ffn = config_dict["use_swiglu_ffn"]
+        patch_size = config_dict.get("patch_size", 14)
+        image_size = config_dict.get("image_size", 518)
+        use_mask_token = config_dict.get("use_mask_token", True)
 
-        self.embeddings = Dino2Embeddings(dim, dtype, device, operations)
-        self.encoder = Dino2Encoder(dim, heads, layer_norm_eps, num_layers, dtype, device, operations, use_swiglu_ffn = use_swiglu_ffn)
+        # DA3 extensions (all default to disabled).
+        self.alt_start = config_dict.get("alt_start", -1)
+        self.qknorm_start = config_dict.get("qknorm_start", -1)
+        self.rope_start = config_dict.get("rope_start", -1)
+        self.cat_token = config_dict.get("cat_token", False)
+        rope_freq = config_dict.get("rope_freq", 100.0)
+
+        self.embed_dim = dim
+        self.patch_size = patch_size
+        self.num_register_tokens = 0
+        self.patch_start_idx = 1
+
+        if self.rope_start != -1 and rope_freq > 0:
+            self.rope = RotaryPositionEmbedding2D(frequency=rope_freq)
+            self._position_getter = _PositionGetter()
+        else:
+            self.rope = None
+            self._position_getter = None
+
+        # camera_token shape: (1, 2, dim) -> (ref_token, src_token).
+        num_cam_tokens = 2 if self.alt_start != -1 else 0
+
+        self.embeddings = Dino2Embeddings(
+            dim, dtype, device, operations,
+            patch_size=patch_size, image_size=image_size,
+            use_mask_token=use_mask_token, num_camera_tokens=num_cam_tokens,
+        )
+        self.encoder = Dino2Encoder(
+            dim, heads, layer_norm_eps, num_layers, dtype, device, operations,
+            use_swiglu_ffn=use_swiglu_ffn,
+            qknorm_start=self.qknorm_start,
+        )
         self.layernorm = operations.LayerNorm(dim, eps=layer_norm_eps, dtype=dtype, device=device)
 
     def forward(self, pixel_values, attention_mask=None, intermediate_output=None):
+        if self.alt_start != -1:
+            raise RuntimeError(
+                "Dinov2Model.forward() is the backward-compatible CLIP-vision path and does not "
+                "apply DA3 extensions (RoPE, alternating attention, camera-token injection). "
+                "Use get_intermediate_layers_da3() for Depth Anything 3 models."
+            )
         x = self.embeddings(pixel_values)
         x, i = self.encoder(x, intermediate_output=intermediate_output)
         x = self.layernorm(x)
@@ -181,6 +348,7 @@ class Dinov2Model(torch.nn.Module):
         return x, i, pooled_output, None
 
     def get_intermediate_layers(self, pixel_values, indices, apply_norm=True):
+        """Single-view multi-layer feature extraction."""
         x = self.embeddings(pixel_values)
         optimized_attention = optimized_attention_for_device(x.device, False, small_input=True)
         n_layers = len(self.encoder.layer)
@@ -197,3 +365,132 @@ class Dinov2Model(torch.nn.Module):
             if i >= max_idx:
                 break
         return [cache[i] for i in resolved]
+
+    # ------------------------------------------------------------------
+    # Depth Anything 3 forward
+    # ------------------------------------------------------------------
+    def _prepare_rope_positions(self, B, S, H, W, device):
+        if self.rope is None:
+            return None, None
+        ph, pw = H // self.patch_size, W // self.patch_size
+        pos = self._position_getter(B * S, ph, pw, device=device)
+        # Shift so the cls/cam token at position 0 is reserved for "no diff".
+        pos = pos + 1
+        cls_pos = torch.zeros(B * S, self.patch_start_idx, 2, device=device, dtype=pos.dtype)
+        # Per-view local: real grid positions for patches, 0 for cls token.
+        pos_local = torch.cat([cls_pos, pos], dim=1)
+        # Global (across views): same grid positions; cls token still at 0,
+        # but patches share the same positions in every view.
+        pos_global = torch.cat([cls_pos, torch.zeros_like(pos) + 1], dim=1)
+        return pos_local, pos_global
+
+    def _inject_camera_token(self, x: torch.Tensor, B: int, S: int, cam_token: "torch.Tensor | None") -> torch.Tensor:
+        # x: (B, S, N, C). Replace token at index 0 with the camera token.
+        if cam_token is not None:
+            inj = cam_token
+        else:
+            ct = comfy.model_management.cast_to_device(self.embeddings.camera_token, x.device, x.dtype)
+            ref_token = ct[:, :1].expand(B, -1, -1)
+            src_token = ct[:, 1:].expand(B, max(S - 1, 0), -1)
+            inj = torch.cat([ref_token, src_token], dim=1)
+        x = x.clone()
+        x[:, :, 0] = inj
+        return x
+
+    def get_intermediate_layers_da3(self, pixel_values, out_layers, cam_token=None, ref_view_strategy="saddle_balanced", export_feat_layers=None):
+        """Multi-view multi-layer feature extraction used by Depth Anything 3."""
+        if pixel_values.ndim == 4:
+            pixel_values = pixel_values.unsqueeze(1)
+        assert pixel_values.ndim == 5 and pixel_values.shape[2] == 3, \
+            f"expected (B,3,H,W) or (B,S,3,H,W); got {tuple(pixel_values.shape)}"
+        B, S, _, H, W = pixel_values.shape
+
+        # Patch + cls + (interpolated) pos embed for each view.
+        x = pixel_values.reshape(B * S, 3, H, W)
+        x = self.embeddings(x)                          # (B*S, 1+N, C)
+        x = x.reshape(B, S, x.shape[-2], x.shape[-1])    # (B, S, 1+N, C)
+
+        pos_local, pos_global = self._prepare_rope_positions(B, S, H, W, x.device)
+        # optimized_attention is only used by blocks without QK-norm/RoPE
+        # (vanilla DINOv2 path); enabling-aware blocks fall through to SDPA.
+        optimized_attention = optimized_attention_for_device(x.device, False, small_input=True)
+
+        out_set = set(out_layers)
+        export_set = set(export_feat_layers) if export_feat_layers else set()
+        outputs: list[torch.Tensor] = []
+        aux_outputs: list[torch.Tensor] = []
+        local_x = x
+        b_idx = None
+
+
+        for i, blk in enumerate(self.encoder.layer):
+            apply_rope = self.rope is not None and i >= self.rope_start
+            block_rope = self.rope if apply_rope else None
+            l_pos = pos_local if apply_rope else None
+            g_pos = pos_global if apply_rope else None
+
+            # Reference-view selection threshold: matches the upstream constant
+            # THRESH_FOR_REF_SELECTION = 3. Skipped when a user-supplied
+            # cam_token is provided (camera info already pins the geometry).
+            if (self.alt_start != -1 and i == self.alt_start - 1 and S >= THRESH_FOR_REF_SELECTION and cam_token is None):
+                b_idx = select_reference_view(x, strategy=ref_view_strategy)
+                x = reorder_by_reference(x, b_idx)
+                local_x = reorder_by_reference(local_x, b_idx)
+
+            if self.alt_start != -1 and i == self.alt_start:
+                x = self._inject_camera_token(x, B, S, cam_token)
+
+            if self.alt_start != -1 and i >= self.alt_start and (i % 2 == 1):
+                # Global attention across views: flatten S into the seq dim.
+                t = x.reshape(B, S * x.shape[-2], x.shape[-1])
+                p = g_pos.reshape(B, S * g_pos.shape[-2], g_pos.shape[-1]) if g_pos is not None else None
+                t = blk(t, optimized_attention=optimized_attention, pos=p, rope=block_rope)
+                x = t.reshape(B, S, x.shape[-2], x.shape[-1])
+            else:
+                # Per-view local attention.
+                t = x.reshape(B * S, x.shape[-2], x.shape[-1])
+                p = l_pos.reshape(B * S, l_pos.shape[-2], l_pos.shape[-1]) if l_pos is not None else None
+                t = blk(t, optimized_attention=optimized_attention, pos=p, rope=block_rope)
+                x = t.reshape(B, S, x.shape[-2], x.shape[-1])
+                local_x = x
+
+            if i in out_set:
+                if self.cat_token:
+                    out_x = torch.cat([local_x, x], dim=-1)
+                else:
+                    out_x = x
+                # Restore original view order on the way out so heads see views
+                # in the user's expected order.
+                if b_idx is not None and self.alt_start != -1:
+                    out_x = restore_original_order(out_x, b_idx)
+                outputs.append(out_x)
+
+            if i in export_set:
+                aux = x
+                if b_idx is not None and self.alt_start != -1:
+                    aux = restore_original_order(aux, b_idx)
+                aux_outputs.append(aux)
+
+        # Apply final norm. When cat_token is set, only the right half
+        # ("global" features) is normalised; the left half is left as-is to
+        # match the upstream DA3 head signature.
+        normed: list[torch.Tensor] = []
+        cls_tokens: list[torch.Tensor] = []
+        for out_x in outputs:
+            cls_tokens.append(out_x[:, :, 0])
+            if out_x.shape[-1] == self.embed_dim:
+                normed.append(self.layernorm(out_x))
+            elif out_x.shape[-1] == self.embed_dim * 2:
+                left = out_x[..., :self.embed_dim]
+                right = self.layernorm(out_x[..., self.embed_dim:])
+                normed.append(torch.cat([left, right], dim=-1))
+            else:
+                raise ValueError(f"Unexpected token width: {out_x.shape[-1]}")
+
+        # Drop cls/cam token from the patch sequence.
+        normed = [o[..., 1 + self.num_register_tokens:, :] for o in normed]
+
+        # Final layernorm + drop cls token from auxiliary features too.
+        aux_normed = [self.layernorm(o)[..., 1 + self.num_register_tokens:, :]
+                      for o in aux_outputs]
+        return list(zip(normed, cls_tokens)), aux_normed

comfy/image_encoders/dino3.py

@@ -0,0 +1,259 @@
+import math
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+import comfy.ops
+from comfy.ldm.modules.attention import optimized_attention_for_device
+from comfy.image_encoders.dino2 import LayerScale as DINOv3ViTLayerScale
+
+
+# DINOv3 ViT-H/16+ (SwiGLU)
+DINOV3_VITH_CONFIG = {
+    "model_type": "dinov3",
+    "num_hidden_layers": 32,
+    "hidden_size": 1280,
+    "num_attention_heads": 20,
+    "num_register_tokens": 4,
+    "intermediate_size": 5120,
+    "layer_norm_eps": 1e-5,
+    "num_channels": 3,
+    "patch_size": 16,
+    "rope_theta": 100.0,
+    "use_gated_mlp": True,
+    "gated_mlp_act": "silu",
+    "image_size": 1024,
+    "image_mean": [0.485, 0.456, 0.406],
+    "image_std": [0.229, 0.224, 0.225],
+}
+
+
+class DINOv3ViTMLP(nn.Module):
+    def __init__(self, hidden_size, intermediate_size, mlp_bias, device, dtype, operations):
+        super().__init__()
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.up_proj = operations.Linear(self.hidden_size, self.intermediate_size, bias=mlp_bias, device=device, dtype=dtype)
+        self.down_proj = operations.Linear(self.intermediate_size, self.hidden_size, bias=mlp_bias, device=device, dtype=dtype)
+        self.act_fn = torch.nn.GELU()
+
+    def forward(self, x):
+        return self.down_proj(self.act_fn(self.up_proj(x)))
+
+
+def rotate_half(x):
+    x1 = x[..., : x.shape[-1] // 2]
+    x2 = x[..., x.shape[-1] // 2 :]
+    return torch.cat((-x2, x1), dim=-1)
+
+
+def apply_rotary_pos_emb(q, k, cos, sin, **kwargs):
+    num_tokens = q.shape[-2]
+    num_patches = sin.shape[-2]
+    num_prefix_tokens = num_tokens - num_patches
+
+    q_prefix_tokens, q_patches = q.split((num_prefix_tokens, num_patches), dim=-2)
+    k_prefix_tokens, k_patches = k.split((num_prefix_tokens, num_patches), dim=-2)
+
+    q_patches = (q_patches * cos) + (rotate_half(q_patches) * sin)
+    k_patches = (k_patches * cos) + (rotate_half(k_patches) * sin)
+
+    q = torch.cat((q_prefix_tokens, q_patches), dim=-2)
+    k = torch.cat((k_prefix_tokens, k_patches), dim=-2)
+
+    return q, k
+
+
+class DINOv3ViTAttention(nn.Module):
+    def __init__(self, hidden_size, num_attention_heads, device, dtype, operations):
+        super().__init__()
+        self.embed_dim = hidden_size
+        self.num_heads = num_attention_heads
+        self.head_dim = self.embed_dim // self.num_heads
+
+        self.k_proj = operations.Linear(self.embed_dim, self.embed_dim, bias=False, device=device, dtype=dtype)  # key_bias = False
+        self.v_proj = operations.Linear(self.embed_dim, self.embed_dim, bias=True, device=device, dtype=dtype)
+        self.q_proj = operations.Linear(self.embed_dim, self.embed_dim, bias=True, device=device, dtype=dtype)
+        self.o_proj = operations.Linear(self.embed_dim, self.embed_dim, bias=True, device=device, dtype=dtype)
+
+    def forward(self, hidden_states, attention_mask=None, position_embeddings=None, **kwargs):
+        batch_size, patches, _ = hidden_states.size()
+
+        query_states = self.q_proj(hidden_states)
+        key_states = self.k_proj(hidden_states)
+        value_states = self.v_proj(hidden_states)
+
+        query_states = query_states.view(batch_size, patches, self.num_heads, self.head_dim).transpose(1, 2)
+        key_states = key_states.view(batch_size, patches, self.num_heads, self.head_dim).transpose(1, 2)
+        value_states = value_states.view(batch_size, patches, self.num_heads, self.head_dim).transpose(1, 2)
+
+        if position_embeddings is not None:
+            cos, sin = position_embeddings
+            query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)
+
+        attn = optimized_attention_for_device(query_states.device, mask=False)
+        attn_output = attn(
+            query_states, key_states, value_states, self.num_heads, attention_mask,
+            skip_reshape=True, skip_output_reshape=True, low_precision_attention=False,
+        )
+
+        attn_output = attn_output.transpose(1, 2)
+        attn_output = attn_output.reshape(batch_size, patches, -1).contiguous()
+        attn_output = self.o_proj(attn_output)
+        return attn_output
+
+
+class DINOv3ViTGatedMLP(nn.Module):
+    def __init__(self, hidden_size, intermediate_size, mlp_bias, device, dtype, operations, act="silu"):
+        super().__init__()
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.gate_proj = operations.Linear(self.hidden_size, self.intermediate_size, bias=mlp_bias, device=device, dtype=dtype)
+        self.up_proj = operations.Linear(self.hidden_size, self.intermediate_size, bias=mlp_bias, device=device, dtype=dtype)
+        self.down_proj = operations.Linear(self.intermediate_size, self.hidden_size, bias=mlp_bias, device=device, dtype=dtype)
+        self.act_fn = torch.nn.SiLU() if act == "silu" else torch.nn.GELU()
+
+    def forward(self, x):
+        return self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))
+
+
+def get_patches_center_coordinates(num_patches_h, num_patches_w, dtype, device):
+    coords_h = torch.arange(0.5, num_patches_h, dtype=dtype, device=device)
+    coords_w = torch.arange(0.5, num_patches_w, dtype=dtype, device=device)
+    coords_h = coords_h / num_patches_h
+    coords_w = coords_w / num_patches_w
+    coords = torch.stack(torch.meshgrid(coords_h, coords_w, indexing="ij"), dim=-1)
+    coords = coords.flatten(0, 1)
+    coords = 2.0 * coords - 1.0
+    return coords
+
+
+class DINOv3ViTRopePositionEmbedding(nn.Module):
+    inv_freq: torch.Tensor
+
+    def __init__(self, rope_theta, hidden_size, num_attention_heads, patch_size, device, dtype):
+        super().__init__()
+        self.base = rope_theta
+        self.head_dim = hidden_size // num_attention_heads
+        self.patch_size = patch_size
+
+        inv_freq = 1 / self.base ** torch.arange(0, 1, 4 / self.head_dim, dtype=torch.float32, device=device)
+        self.register_buffer("inv_freq", inv_freq, persistent=False)
+
+    def forward(self, pixel_values):
+        _, _, height, width = pixel_values.shape
+        num_patches_h = height // self.patch_size
+        num_patches_w = width // self.patch_size
+
+        patch_coords = get_patches_center_coordinates(num_patches_h, num_patches_w, dtype=torch.float32, device=pixel_values.device)
+        self.inv_freq = self.inv_freq.to(pixel_values.device)
+        angles = 2 * math.pi * patch_coords[:, :, None] * self.inv_freq[None, None, :]
+        angles = angles.flatten(1, 2)
+        angles = angles.tile(2)
+        cos = torch.cos(angles).to(dtype=pixel_values.dtype)
+        sin = torch.sin(angles).to(dtype=pixel_values.dtype)
+        return cos, sin
+
+
+class DINOv3ViTEmbeddings(nn.Module):
+    def __init__(self, hidden_size, num_register_tokens, num_channels, patch_size, dtype, device, operations):
+        super().__init__()
+        self.cls_token = nn.Parameter(torch.empty(1, 1, hidden_size, device=device, dtype=dtype))
+        self.mask_token = nn.Parameter(torch.empty(1, 1, hidden_size, device=device, dtype=dtype))
+        self.register_tokens = nn.Parameter(torch.empty(1, num_register_tokens, hidden_size, device=device, dtype=dtype))
+        self.patch_embeddings = operations.Conv2d(
+            num_channels, hidden_size, kernel_size=patch_size, stride=patch_size, device=device, dtype=dtype
+        )
+
+    def forward(self, pixel_values, bool_masked_pos=None):
+        batch_size = pixel_values.shape[0]
+
+        patch_embeddings = self.patch_embeddings(pixel_values)
+        patch_embeddings = patch_embeddings.flatten(2).transpose(1, 2)
+
+        if bool_masked_pos is not None:
+            mask_token = comfy.ops.cast_to_input(self.mask_token, patch_embeddings)
+            patch_embeddings = torch.where(bool_masked_pos.unsqueeze(-1), mask_token, patch_embeddings)
+
+        cls_token = comfy.ops.cast_to_input(self.cls_token.expand(batch_size, -1, -1), patch_embeddings)
+        register_tokens = comfy.ops.cast_to_input(self.register_tokens.expand(batch_size, -1, -1), patch_embeddings)
+        embeddings = torch.cat([cls_token, register_tokens, patch_embeddings], dim=1)
+        return embeddings
+
+
+class DINOv3ViTLayer(nn.Module):
+    def __init__(self, hidden_size, layer_norm_eps, use_gated_mlp, mlp_bias, intermediate_size,
+                 num_attention_heads, device, dtype, operations, gated_mlp_act="silu"):
+        super().__init__()
+        self.norm1 = operations.LayerNorm(hidden_size, eps=layer_norm_eps, device=device, dtype=dtype)
+        self.attention = DINOv3ViTAttention(hidden_size, num_attention_heads, device=device, dtype=dtype, operations=operations)
+        self.layer_scale1 = DINOv3ViTLayerScale(hidden_size, device=device, dtype=dtype, operations=None)
+
+        self.norm2 = operations.LayerNorm(hidden_size, eps=layer_norm_eps, device=device, dtype=dtype)
+        if use_gated_mlp:
+            self.mlp = DINOv3ViTGatedMLP(hidden_size, intermediate_size, mlp_bias, device=device, dtype=dtype, operations=operations, act=gated_mlp_act)
+        else:
+            self.mlp = DINOv3ViTMLP(hidden_size, intermediate_size=intermediate_size, mlp_bias=mlp_bias, device=device, dtype=dtype, operations=operations)
+        self.layer_scale2 = DINOv3ViTLayerScale(hidden_size, device=device, dtype=dtype, operations=None)
+
+    def forward(self, hidden_states, attention_mask=None, position_embeddings=None):
+        residual = hidden_states
+        hidden_states = self.norm1(hidden_states)
+        hidden_states = self.attention(hidden_states, attention_mask=attention_mask, position_embeddings=position_embeddings)
+        hidden_states = self.layer_scale1(hidden_states)
+        hidden_states = hidden_states + residual
+
+        residual = hidden_states
+        hidden_states = self.norm2(hidden_states)
+        hidden_states = self.mlp(hidden_states)
+        hidden_states = self.layer_scale2(hidden_states)
+        hidden_states = hidden_states + residual
+        return hidden_states
+
+
+class DINOv3ViTModel(nn.Module):
+    def __init__(self, config, dtype, device, operations):
+        super().__init__()
+        num_hidden_layers = config["num_hidden_layers"]
+        hidden_size = config["hidden_size"]
+        num_attention_heads = config["num_attention_heads"]
+        num_register_tokens = config["num_register_tokens"]
+        intermediate_size = config["intermediate_size"]
+        layer_norm_eps = config["layer_norm_eps"]
+        num_channels = config["num_channels"]
+        patch_size = config["patch_size"]
+        rope_theta = config["rope_theta"]
+        use_gated_mlp = config.get("use_gated_mlp", False)
+        gated_mlp_act = config.get("gated_mlp_act", "silu")
+
+        self.embeddings = DINOv3ViTEmbeddings(
+            hidden_size, num_register_tokens, num_channels=num_channels, patch_size=patch_size,
+            dtype=dtype, device=device, operations=operations
+        )
+        self.rope_embeddings = DINOv3ViTRopePositionEmbedding(
+            rope_theta, hidden_size, num_attention_heads, patch_size=patch_size, dtype=dtype, device=device
+        )
+        self.layer = nn.ModuleList([
+            DINOv3ViTLayer(hidden_size, layer_norm_eps, use_gated_mlp=use_gated_mlp, mlp_bias=True,
+                           intermediate_size=intermediate_size, num_attention_heads=num_attention_heads,
+                           dtype=dtype, device=device, operations=operations, gated_mlp_act=gated_mlp_act)
+            for _ in range(num_hidden_layers)])
+        self.norm = operations.LayerNorm(hidden_size, eps=layer_norm_eps, dtype=dtype, device=device)
+
+    def get_input_embeddings(self):
+        return self.embeddings.patch_embeddings
+
+    def forward(self, pixel_values, bool_masked_pos=None, **kwargs):
+        hidden_states = self.embeddings(pixel_values, bool_masked_pos=bool_masked_pos)
+        position_embeddings = self.rope_embeddings(pixel_values)
+
+        for layer_module in self.layer:
+            hidden_states = layer_module(hidden_states, position_embeddings=position_embeddings)
+
+        if kwargs.get("skip_norm_elementwise", False):
+            sequence_output = F.layer_norm(hidden_states, hidden_states.shape[-1:])
+        else:
+            norm = self.norm.to(hidden_states.device)
+            sequence_output = norm(hidden_states)
+        pooled_output = sequence_output[:, 0, :]
+        return sequence_output, None, pooled_output, None

comfy/latent_formats.py

@@ -239,6 +239,16 @@ class Flux2(LatentFormat):
     def process_out(self, latent):
         return latent
 
+class TripoSplat(LatentFormat):
+    # Sequence latent (B, 8192, 16) the camera token rides alongside as a second nested latent
+    latent_channels = 16
+
+    def process_in(self, latent):
+        return latent
+
+    def process_out(self, latent):
+        return latent
+
 class Mochi(LatentFormat):
     latent_channels = 12
     latent_dimensions = 3

comfy/ldm/boogu/model.py

@@ -0,0 +1,321 @@
+# Boogu-Image-0.1 transformer
+# Architecture is an OmniGen2 derivative (see comfy/ldm/omnigen/omnigen2.py) with an
+# added dual-stream ("double_stream") stage before the single-stream layers, conditioned
+# by a Qwen3-VL multimodal LLM. Reuses the OmniGen2/Lumina building blocks and the Flux
+# RoPE core, the only new component is the double-stream block + the hybrid forward order.
+
+from typing import Optional, Tuple
+
+import torch
+import torch.nn as nn
+from einops import rearrange
+
+import comfy.ldm.common_dit
+import comfy.ldm.omnigen.omnigen2
+from comfy.ldm.modules.attention import optimized_attention_masked
+from comfy.ldm.omnigen.omnigen2 import (
+    OmniGen2RotaryPosEmbed,
+    Lumina2CombinedTimestepCaptionEmbedding,
+    LuminaRMSNormZero,
+    LuminaLayerNormContinuous,
+    LuminaFeedForward,
+    Attention,
+    OmniGen2TransformerBlock,
+    apply_rotary_emb,
+)
+
+class BooguDoubleStreamProcessor(nn.Module):
+    # Joint attention over [instruct ; img] with separate per-stream q/k/v and output projections.
+    def __init__(self, dim, head_dim, heads, kv_heads, dtype=None, device=None, operations=None):
+        super().__init__()
+        query_dim = head_dim * heads
+        kv_dim = head_dim * kv_heads
+
+        self.img_to_q = operations.Linear(query_dim, query_dim, bias=False, dtype=dtype, device=device)
+        self.img_to_k = operations.Linear(query_dim, kv_dim, bias=False, dtype=dtype, device=device)
+        self.img_to_v = operations.Linear(query_dim, kv_dim, bias=False, dtype=dtype, device=device)
+
+        self.instruct_to_q = operations.Linear(query_dim, query_dim, bias=False, dtype=dtype, device=device)
+        self.instruct_to_k = operations.Linear(query_dim, kv_dim, bias=False, dtype=dtype, device=device)
+        self.instruct_to_v = operations.Linear(query_dim, kv_dim, bias=False, dtype=dtype, device=device)
+
+        self.instruct_out = operations.Linear(query_dim, query_dim, bias=False, dtype=dtype, device=device)
+        self.img_out = operations.Linear(query_dim, query_dim, bias=False, dtype=dtype, device=device)
+
+    def forward(self, attn, img_hidden_states, instruct_hidden_states, rotary_emb, attention_mask=None, transformer_options={}):
+        batch_size = img_hidden_states.shape[0]
+        L_instruct = instruct_hidden_states.shape[1]
+
+        img_q = self.img_to_q(img_hidden_states)
+        img_k = self.img_to_k(img_hidden_states)
+        img_v = self.img_to_v(img_hidden_states)
+
+        instruct_q = self.instruct_to_q(instruct_hidden_states)
+        instruct_k = self.instruct_to_k(instruct_hidden_states)
+        instruct_v = self.instruct_to_v(instruct_hidden_states)
+
+        # Concatenate instruction first, then image (matches reference processor order).
+        query = torch.cat([instruct_q, img_q], dim=1)
+        key = torch.cat([instruct_k, img_k], dim=1)
+        value = torch.cat([instruct_v, img_v], dim=1)
+
+        query = query.view(batch_size, -1, attn.heads, attn.dim_head)
+        key = key.view(batch_size, -1, attn.kv_heads, attn.dim_head)
+        value = value.view(batch_size, -1, attn.kv_heads, attn.dim_head)
+
+        query = attn.norm_q(query)
+        key = attn.norm_k(key)
+
+        if rotary_emb is not None:
+            query = apply_rotary_emb(query, rotary_emb)
+            key = apply_rotary_emb(key, rotary_emb)
+
+        query = query.transpose(1, 2)
+        key = key.transpose(1, 2)
+        value = value.transpose(1, 2)
+
+        if attn.kv_heads < attn.heads:
+            key = key.repeat_interleave(attn.heads // attn.kv_heads, dim=1)
+            value = value.repeat_interleave(attn.heads // attn.kv_heads, dim=1)
+
+        hidden_states = optimized_attention_masked(query, key, value, attn.heads, attention_mask, skip_reshape=True, transformer_options=transformer_options)
+
+        # Split back to instruction/image, apply per-stream output projections, recombine.
+        instruct_hidden_states = self.instruct_out(hidden_states[:, :L_instruct])
+        img_hidden_states = self.img_out(hidden_states[:, L_instruct:])
+        hidden_states = torch.cat([instruct_hidden_states, img_hidden_states], dim=1)
+
+        hidden_states = attn.to_out[0](hidden_states)
+        return hidden_states
+
+
+class BooguJointAttention(nn.Module):
+    # Holds the shared q/k RMSNorm + final output projection
+    def __init__(self, dim, head_dim, heads, kv_heads, eps=1e-5, dtype=None, device=None, operations=None):
+        super().__init__()
+        self.heads = heads
+        self.kv_heads = kv_heads
+        self.dim_head = head_dim
+        self.scale = head_dim ** -0.5
+
+        self.norm_q = operations.RMSNorm(head_dim, eps=eps, dtype=dtype, device=device)
+        self.norm_k = operations.RMSNorm(head_dim, eps=eps, dtype=dtype, device=device)
+        self.to_out = nn.Sequential(
+            operations.Linear(heads * head_dim, dim, bias=False, dtype=dtype, device=device),
+            nn.Dropout(0.0),
+        )
+        self.processor = BooguDoubleStreamProcessor(dim, head_dim, heads, kv_heads, dtype=dtype, device=device, operations=operations)
+
+    def forward(self, img_hidden_states, instruct_hidden_states, rotary_emb, attention_mask=None, transformer_options={}):
+        return self.processor(self, img_hidden_states, instruct_hidden_states, rotary_emb, attention_mask, transformer_options=transformer_options)
+
+
+class BooguDoubleStreamBlock(nn.Module):
+    # Dual-stream block: joint attention over [instruct ; img] + image self-attention, each stream with its own modulation/MLP.
+    def __init__(self, dim, num_attention_heads, num_kv_heads, multiple_of, ffn_dim_multiplier, norm_eps, dtype=None, device=None, operations=None):
+        super().__init__()
+        head_dim = dim // num_attention_heads
+
+        self.img_instruct_attn = BooguJointAttention(dim, head_dim, num_attention_heads, num_kv_heads, eps=1e-5, dtype=dtype, device=device, operations=operations)
+        self.img_self_attn = Attention(
+            query_dim=dim, dim_head=head_dim, heads=num_attention_heads, kv_heads=num_kv_heads,
+            eps=1e-5, bias=False, dtype=dtype, device=device, operations=operations,
+        )
+
+        self.img_feed_forward = LuminaFeedForward(dim=dim, inner_dim=4 * dim, multiple_of=multiple_of, dtype=dtype, device=device, operations=operations)
+        self.instruct_feed_forward = LuminaFeedForward(dim=dim, inner_dim=4 * dim, multiple_of=multiple_of, dtype=dtype, device=device, operations=operations)
+
+        self.img_norm1 = LuminaRMSNormZero(embedding_dim=dim, norm_eps=norm_eps, dtype=dtype, device=device, operations=operations)
+        self.img_norm2 = LuminaRMSNormZero(embedding_dim=dim, norm_eps=norm_eps, dtype=dtype, device=device, operations=operations)
+        self.img_norm3 = LuminaRMSNormZero(embedding_dim=dim, norm_eps=norm_eps, dtype=dtype, device=device, operations=operations)
+        self.instruct_norm1 = LuminaRMSNormZero(embedding_dim=dim, norm_eps=norm_eps, dtype=dtype, device=device, operations=operations)
+        self.instruct_norm2 = LuminaRMSNormZero(embedding_dim=dim, norm_eps=norm_eps, dtype=dtype, device=device, operations=operations)
+
+        self.img_attn_norm = operations.RMSNorm(dim, eps=norm_eps, dtype=dtype, device=device)
+        self.img_self_attn_norm = operations.RMSNorm(dim, eps=norm_eps, dtype=dtype, device=device)
+        self.img_ffn_norm1 = operations.RMSNorm(dim, eps=norm_eps, dtype=dtype, device=device)
+        self.img_ffn_norm2 = operations.RMSNorm(dim, eps=norm_eps, dtype=dtype, device=device)
+
+        self.instruct_attn_norm = operations.RMSNorm(dim, eps=norm_eps, dtype=dtype, device=device)
+        self.instruct_ffn_norm1 = operations.RMSNorm(dim, eps=norm_eps, dtype=dtype, device=device)
+        self.instruct_ffn_norm2 = operations.RMSNorm(dim, eps=norm_eps, dtype=dtype, device=device)
+
+    def forward(self, img_hidden_states, instruct_hidden_states, joint_rotary_emb, img_rotary_emb, temb, joint_attention_mask=None, img_attention_mask=None, transformer_options={}):
+        L_instruct = instruct_hidden_states.shape[1]
+
+        img_norm1_out, img_gate_msa, img_scale_mlp, img_gate_mlp = self.img_norm1(img_hidden_states, temb)
+        img_norm2_out, img_shift_mlp, _, _ = self.img_norm2(img_hidden_states, temb)
+        img_norm3_out, img_gate_self, _, _ = self.img_norm3(img_hidden_states, temb)
+
+        instruct_norm1_out, instruct_gate_msa, instruct_scale_mlp, instruct_gate_mlp = self.instruct_norm1(instruct_hidden_states, temb)
+        instruct_norm2_out, instruct_shift_mlp, _, _ = self.instruct_norm2(instruct_hidden_states, temb)
+
+        joint_attn_out = self.img_instruct_attn(img_norm1_out, instruct_norm1_out, joint_rotary_emb, joint_attention_mask, transformer_options=transformer_options)
+        instruct_attn_out = joint_attn_out[:, :L_instruct]
+        img_attn_out = joint_attn_out[:, L_instruct:]
+
+        img_self_attn_out = self.img_self_attn(img_norm3_out, img_norm3_out, img_attention_mask, img_rotary_emb, transformer_options=transformer_options)
+
+        img_hidden_states = img_hidden_states + img_gate_msa.unsqueeze(1).tanh() * self.img_attn_norm(img_attn_out)
+        img_hidden_states = img_hidden_states + img_gate_self.unsqueeze(1).tanh() * self.img_self_attn_norm(img_self_attn_out)
+        img_mlp_input = (1 + img_scale_mlp.unsqueeze(1)) * img_norm2_out + img_shift_mlp.unsqueeze(1)
+        img_mlp_out = self.img_feed_forward(self.img_ffn_norm1(img_mlp_input))
+        img_hidden_states = img_hidden_states + img_gate_mlp.unsqueeze(1).tanh() * self.img_ffn_norm2(img_mlp_out)
+
+        instruct_hidden_states = instruct_hidden_states + instruct_gate_msa.unsqueeze(1).tanh() * self.instruct_attn_norm(instruct_attn_out)
+        instruct_mlp_input = (1 + instruct_scale_mlp.unsqueeze(1)) * instruct_norm2_out + instruct_shift_mlp.unsqueeze(1)
+        instruct_mlp_out = self.instruct_feed_forward(self.instruct_ffn_norm1(instruct_mlp_input))
+        instruct_hidden_states = instruct_hidden_states + instruct_gate_mlp.unsqueeze(1).tanh() * self.instruct_ffn_norm2(instruct_mlp_out)
+
+        return img_hidden_states, instruct_hidden_states
+
+
+class BooguTransformer2DModel(nn.Module):
+    def __init__(
+        self,
+        patch_size: int = 2,
+        in_channels: int = 16,
+        out_channels: Optional[int] = None,
+        hidden_size: int = 3360,
+        num_layers: int = 32,
+        num_double_stream_layers: int = 8,
+        num_refiner_layers: int = 2,
+        num_attention_heads: int = 28,
+        num_kv_heads: int = 7,
+        multiple_of: int = 256,
+        ffn_dim_multiplier: Optional[float] = None,
+        norm_eps: float = 1e-5,
+        axes_dim_rope: Tuple[int, int, int] = (40, 40, 40),
+        axes_lens: Tuple[int, int, int] = (2048, 1664, 1664),
+        instruction_feat_dim: int = 4096,
+        timestep_scale: float = 1000.0,
+        image_model=None,
+        device=None, dtype=None, operations=None,
+    ):
+        super().__init__()
+
+        self.patch_size = patch_size
+        self.out_channels = out_channels or in_channels
+        self.hidden_size = hidden_size
+        self.dtype = dtype
+
+        self.rope_embedder = OmniGen2RotaryPosEmbed(
+            theta=10000,
+            axes_dim=axes_dim_rope,
+            axes_lens=axes_lens,
+            patch_size=patch_size,
+        )
+
+        self.x_embedder = operations.Linear(patch_size * patch_size * in_channels, hidden_size, dtype=dtype, device=device)
+        self.ref_image_patch_embedder = operations.Linear(patch_size * patch_size * in_channels, hidden_size, dtype=dtype, device=device)
+
+        self.time_caption_embed = Lumina2CombinedTimestepCaptionEmbedding(
+            hidden_size=hidden_size,
+            text_feat_dim=instruction_feat_dim,
+            norm_eps=norm_eps,
+            timestep_scale=timestep_scale, dtype=dtype, device=device, operations=operations
+        )
+
+        self.noise_refiner = nn.ModuleList([
+            OmniGen2TransformerBlock(hidden_size, num_attention_heads, num_kv_heads, multiple_of, ffn_dim_multiplier, norm_eps, modulation=True, dtype=dtype, device=device, operations=operations)
+            for _ in range(num_refiner_layers)
+        ])
+
+        self.ref_image_refiner = nn.ModuleList([
+            OmniGen2TransformerBlock(hidden_size, num_attention_heads, num_kv_heads, multiple_of, ffn_dim_multiplier, norm_eps, modulation=True, dtype=dtype, device=device, operations=operations)
+            for _ in range(num_refiner_layers)
+        ])
+
+        self.context_refiner = nn.ModuleList([
+            OmniGen2TransformerBlock(hidden_size, num_attention_heads, num_kv_heads, multiple_of, ffn_dim_multiplier, norm_eps, modulation=False, dtype=dtype, device=device, operations=operations)
+            for _ in range(num_refiner_layers)
+        ])
+
+        self.double_stream_layers = nn.ModuleList([
+            BooguDoubleStreamBlock(hidden_size, num_attention_heads, num_kv_heads, multiple_of, ffn_dim_multiplier, norm_eps, dtype=dtype, device=device, operations=operations)
+            for _ in range(num_double_stream_layers)
+        ])
+
+        self.single_stream_layers = nn.ModuleList([
+            OmniGen2TransformerBlock(hidden_size, num_attention_heads, num_kv_heads, multiple_of, ffn_dim_multiplier, norm_eps, modulation=True, dtype=dtype, device=device, operations=operations)
+            for _ in range(num_layers)
+        ])
+
+        self.norm_out = LuminaLayerNormContinuous(
+            embedding_dim=hidden_size,
+            conditioning_embedding_dim=min(hidden_size, 1024),
+            elementwise_affine=False,
+            eps=1e-6,
+            out_dim=patch_size * patch_size * self.out_channels, dtype=dtype, device=device, operations=operations
+        )
+
+        self.image_index_embedding = nn.Parameter(torch.empty(5, hidden_size, device=device, dtype=dtype))
+
+    # Patchify/refine helpers are identical to OmniGen2; reuse via bound methods.
+    flat_and_pad_to_seq = comfy.ldm.omnigen.omnigen2.OmniGen2Transformer2DModel.flat_and_pad_to_seq
+    img_patch_embed_and_refine = comfy.ldm.omnigen.omnigen2.OmniGen2Transformer2DModel.img_patch_embed_and_refine
+
+    def forward(self, x, timesteps, context, num_tokens, ref_latents=None, attention_mask=None, transformer_options={}, **kwargs):
+        B, C, H, W = x.shape
+        hidden_states = comfy.ldm.common_dit.pad_to_patch_size(x, (self.patch_size, self.patch_size))
+        _, _, H_padded, W_padded = hidden_states.shape
+        timestep = 1.0 - timesteps
+        text_hidden_states = context
+        text_attention_mask = attention_mask
+        ref_image_hidden_states = ref_latents
+        device = hidden_states.device
+
+        temb, text_hidden_states = self.time_caption_embed(timestep, text_hidden_states, hidden_states[0].dtype)
+
+        (
+            hidden_states, ref_image_hidden_states,
+            img_mask, ref_img_mask,
+            l_effective_ref_img_len, l_effective_img_len,
+            ref_img_sizes, img_sizes,
+        ) = self.flat_and_pad_to_seq(hidden_states, ref_image_hidden_states)
+
+        (
+            context_rotary_emb, ref_img_rotary_emb, noise_rotary_emb,
+            rotary_emb, encoder_seq_lengths, seq_lengths,
+        ) = self.rope_embedder(
+            hidden_states.shape[0], text_hidden_states.shape[1], [num_tokens] * text_hidden_states.shape[0],
+            l_effective_ref_img_len, l_effective_img_len,
+            ref_img_sizes, img_sizes, device,
+        )
+
+        for layer in self.context_refiner:
+            text_hidden_states = layer(text_hidden_states, text_attention_mask, context_rotary_emb, transformer_options=transformer_options)
+
+        img_len = hidden_states.shape[1]
+        combined_img_hidden_states = self.img_patch_embed_and_refine(
+            hidden_states, ref_image_hidden_states,
+            img_mask, ref_img_mask,
+            noise_rotary_emb, ref_img_rotary_emb,
+            l_effective_ref_img_len, l_effective_img_len,
+            temb,
+            transformer_options=transformer_options,
+        )
+
+        # Double-stream stage: the image self-attention only sees the [ref ; noise] tokens,
+        # which sit after the instruction tokens in the joint rope.
+        L_instruct = text_hidden_states.shape[1]
+        combined_img_rotary_emb = rotary_emb[:, L_instruct:]
+        for layer in self.double_stream_layers:
+            combined_img_hidden_states, text_hidden_states = layer(
+                combined_img_hidden_states, text_hidden_states,
+                rotary_emb, combined_img_rotary_emb, temb,
+                joint_attention_mask=None, img_attention_mask=None,
+                transformer_options=transformer_options,
+            )
+
+        hidden_states = torch.cat([text_hidden_states, combined_img_hidden_states], dim=1)
+
+        for layer in self.single_stream_layers:
+            hidden_states = layer(hidden_states, None, rotary_emb, temb, transformer_options=transformer_options)
+
+        hidden_states = self.norm_out(hidden_states, temb)
+
+        p = self.patch_size
+        output = rearrange(hidden_states[:, -img_len:], 'b (h w) (p1 p2 c) -> b c (h p1) (w p2)', h=H_padded // p, w=W_padded // p, p1=p, p2=p)[:, :, :H, :W]
+
+        return -output

comfy/ldm/chroma_radiance/model.py

@@ -38,6 +38,8 @@ class ChromaRadianceParams(ChromaParams):
     # None means use the same dtype as the model.
     nerf_embedder_dtype: Optional[torch.dtype]
     use_x0: bool
+    # Use sequential txt_ids instead of zeros
+    use_sequential_txt_ids: bool
 
 class ChromaRadiance(Chroma):
     """
@@ -162,6 +164,9 @@ class ChromaRadiance(Chroma):
         if params.use_x0:
             self.register_buffer("__x0__", torch.tensor([]))
 
+        if params.use_sequential_txt_ids:
+            self.register_buffer("__sequential__", torch.tensor([]))
+
     @property
     def _nerf_final_layer(self) -> nn.Module:
         if self.params.nerf_final_head_type == "linear":
@@ -313,6 +318,9 @@ class ChromaRadiance(Chroma):
         img_ids[:, :, 2] = img_ids[:, :, 2] + torch.linspace(0, w_len - 1, steps=w_len, device=x.device, dtype=x.dtype).unsqueeze(0)
         img_ids = repeat(img_ids, "h w c -> b (h w) c", b=bs)
         txt_ids = torch.zeros((bs, context.shape[1], 3), device=x.device, dtype=x.dtype)
+        # Radiance after 2026-05-22 uses sequential txt_ids instead of zeros
+        if params.use_sequential_txt_ids:
+            txt_ids[:, :, 0] = torch.arange(context.shape[1], device=x.device, dtype=x.dtype).unsqueeze(0).expand(bs, -1)
 
         img_out = self.forward_orig(
             img,

comfy/ldm/colormap.py

@@ -0,0 +1,25 @@
+"""Colormap utilities for depth and geometry visualisation."""
+
+from __future__ import annotations
+
+import torch
+
+
+def turbo(x: torch.Tensor) -> torch.Tensor:
+    """Anton Mikhailov polynomial approximation of the Turbo colormap.
+
+    Args:
+        x: Float tensor with values in [0, 1].
+
+    Returns:
+        RGB tensor of the same shape as ``x`` with a trailing size-3 dimension.
+    """
+    x = x.clamp(0.0, 1.0)
+    x2 = x * x
+    x3 = x2 * x
+    x4 = x2 * x2
+    x5 = x4 * x
+    r = 0.13572138 + 4.61539260*x - 42.66032258*x2 + 132.13108234*x3 - 152.94239396*x4 + 59.28637943*x5
+    g = 0.09140261 + 2.19418839*x + 4.84296658*x2 - 14.18503333*x3 +   4.27729857*x4 +  2.82956604*x5
+    b = 0.10667330 + 12.64194608*x - 60.58204836*x2 + 110.36276771*x3 - 89.90310912*x4 + 27.34824973*x5
+    return torch.stack([r, g, b], dim=-1).clamp(0.0, 1.0)

comfy/ldm/cosmos/predict2.py

@@ -14,6 +14,7 @@ from torchvision import transforms
 import comfy.patcher_extension
 from comfy.ldm.modules.attention import optimized_attention
 import comfy.ldm.common_dit
+import comfy.quant_ops
 
 
 # ---------------------- Feed Forward Network -----------------------
@@ -514,7 +515,7 @@ class Block(nn.Module):
             h=H,
             w=W,
         )
-        x_B_T_H_W_D = x_B_T_H_W_D + gate_self_attn_B_T_1_1_D.to(residual_dtype) * result_B_T_H_W_D.to(residual_dtype)
+        x_B_T_H_W_D = torch.addcmul(x_B_T_H_W_D, gate_self_attn_B_T_1_1_D.to(residual_dtype), result_B_T_H_W_D.to(residual_dtype))
 
         def _x_fn(
             _x_B_T_H_W_D: torch.Tensor,
@@ -547,7 +548,7 @@ class Block(nn.Module):
             shift_cross_attn_B_T_1_1_D,
             transformer_options=transformer_options,
         )
-        x_B_T_H_W_D = result_B_T_H_W_D.to(residual_dtype) * gate_cross_attn_B_T_1_1_D.to(residual_dtype) + x_B_T_H_W_D
+        x_B_T_H_W_D = torch.addcmul(x_B_T_H_W_D, gate_cross_attn_B_T_1_1_D.to(residual_dtype), result_B_T_H_W_D.to(residual_dtype))
 
         normalized_x_B_T_H_W_D = _fn(
             x_B_T_H_W_D,
@@ -556,7 +557,7 @@ class Block(nn.Module):
             shift_mlp_B_T_1_1_D,
         )
         result_B_T_H_W_D = self.mlp(normalized_x_B_T_H_W_D.to(compute_dtype))
-        x_B_T_H_W_D = x_B_T_H_W_D + gate_mlp_B_T_1_1_D.to(residual_dtype) * result_B_T_H_W_D.to(residual_dtype)
+        x_B_T_H_W_D = torch.addcmul(x_B_T_H_W_D, gate_mlp_B_T_1_1_D.to(residual_dtype), result_B_T_H_W_D.to(residual_dtype))
         return x_B_T_H_W_D
 
 

comfy/ldm/depth_anything_3/camera.py

@@ -0,0 +1,177 @@
+"""Camera-token encoder and decoder for Depth Anything 3."""
+
+from __future__ import annotations
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+from comfy.ldm.modules.attention import optimized_attention_for_device
+from .transform import affine_inverse, extri_intri_to_pose_encoding
+
+
+# -----------------------------------------------------------------------
+# Building blocks (mirror depth_anything_3.model.utils.{attention,block})
+# -----------------------------------------------------------------------
+
+
+class _Mlp(nn.Module):
+    """Standard 2-layer MLP with GELU. Matches upstream ``utils.attention.Mlp``."""
+
+    def __init__(self, in_features, hidden_features=None, out_features=None, *, device=None, dtype=None, operations=None):
+        super().__init__()
+        out_features = out_features or in_features
+        hidden_features = hidden_features or in_features
+        self.fc1 = operations.Linear(in_features, hidden_features, bias=True, device=device, dtype=dtype)
+        self.fc2 = operations.Linear(hidden_features, out_features, bias=True, device=device, dtype=dtype)
+
+    def forward(self, x):
+        return self.fc2(F.gelu(self.fc1(x)))
+
+
+class _LayerScale(nn.Module):
+    """Per-channel learnable scaling. Matches upstream LayerScale."""
+
+    def __init__(self, dim, *, device=None, dtype=None):
+        super().__init__()
+        self.gamma = nn.Parameter(torch.empty(dim, device=device, dtype=dtype))
+
+    def forward(self, x):
+        return x * self.gamma.to(dtype=x.dtype, device=x.device)
+
+
+class _Attention(nn.Module):
+    """ Self-attention with fused QKV projection. Mirrors upstream utils.attention.Attention;
+    Layout matches the HF safetensors (attn.qkv.{weight,bias} and attn.proj.{weight,bias})."""
+
+    def __init__(self, dim, num_heads, *, device=None, dtype=None, operations=None):
+        super().__init__()
+        assert dim % num_heads == 0
+        self.num_heads = num_heads
+        self.head_dim = dim // num_heads
+        self.qkv = operations.Linear(dim, dim * 3, bias=True, device=device, dtype=dtype)
+        self.proj = operations.Linear(dim, dim, bias=True, device=device, dtype=dtype)
+
+    def forward(self, x):
+        B, N, C = x.shape
+        qkv = self.qkv(x).reshape(B, N, 3, C)
+        q, k, v = qkv.unbind(2)                      # each (B, N, C)
+        attn_fn = optimized_attention_for_device(x.device, small_input=True)
+        out = attn_fn(q, k, v, heads=self.num_heads)
+        return self.proj(out)
+
+
+class _Block(nn.Module):
+    """Pre-norm transformer block with LayerScale. Used by :class:CameraEnc. Layout follows upstream utils.block.Block."""
+
+    def __init__(self, dim, num_heads, mlp_ratio=4, init_values=0.01, *, device=None, dtype=None, operations=None):
+        super().__init__()
+        self.norm1 = operations.LayerNorm(dim, device=device, dtype=dtype)
+        self.attn = _Attention(dim, num_heads, device=device, dtype=dtype, operations=operations)
+        self.ls1 = _LayerScale(dim, device=device, dtype=dtype) if init_values else nn.Identity()
+        self.norm2 = operations.LayerNorm(dim, device=device, dtype=dtype)
+        self.mlp = _Mlp(in_features=dim, hidden_features=int(dim * mlp_ratio), device=device, dtype=dtype, operations=operations)
+        self.ls2 = _LayerScale(dim, device=device, dtype=dtype) if init_values else nn.Identity()
+
+    def forward(self, x):
+        x = x + self.ls1(self.attn(self.norm1(x)))
+        x = x + self.ls2(self.mlp(self.norm2(x)))
+        return x
+
+
+class CameraEnc(nn.Module):
+    """Encode per-view (extrinsics, intrinsics) into a camera token.
+
+    Maps a 9-D pose-encoding vector through a small MLP up to the backbone's
+    ``embed_dim``, then runs ``trunk_depth`` transformer blocks. The output
+    has shape ``(B, S, embed_dim)`` and is injected at block ``alt_start``
+    of the DINOv2 backbone in place of the cls token.
+
+    Parameters mirror the upstream ``cam_enc.py`` so HF weights load directly.
+    """
+
+    def __init__(
+        self,
+        dim_out: int = 1024,
+        dim_in: int = 9,
+        trunk_depth: int = 4,
+        target_dim: int = 9,
+        num_heads: int = 16,
+        mlp_ratio: int = 4,
+        init_values: float = 0.01,
+        *,
+        device=None, dtype=None, operations=None,
+        **_kwargs,
+    ):
+        super().__init__()
+        self.target_dim = target_dim
+        self.trunk_depth = trunk_depth
+        self.trunk = nn.Sequential(*[
+            _Block(dim_out, num_heads=num_heads, mlp_ratio=mlp_ratio,
+                   init_values=init_values,
+                   device=device, dtype=dtype, operations=operations)
+            for _ in range(trunk_depth)
+        ])
+        self.token_norm = operations.LayerNorm(dim_out, device=device, dtype=dtype)
+        self.trunk_norm = operations.LayerNorm(dim_out, device=device, dtype=dtype)
+        self.pose_branch = _Mlp(
+            in_features=dim_in,
+            hidden_features=dim_out // 2,
+            out_features=dim_out,
+            device=device, dtype=dtype, operations=operations,
+        )
+
+    def forward(self, extrinsics: torch.Tensor, intrinsics: torch.Tensor,
+                image_size_hw) -> torch.Tensor:
+        """Encode camera parameters into ``(B, S, dim_out)`` tokens."""
+        c2ws = affine_inverse(extrinsics)
+        pose_encoding = extri_intri_to_pose_encoding(c2ws, intrinsics, image_size_hw)
+        tokens = self.pose_branch(pose_encoding.to(self.pose_branch.fc1.weight.dtype))
+        tokens = self.token_norm(tokens)
+        tokens = self.trunk(tokens)
+        tokens = self.trunk_norm(tokens)
+        return tokens
+
+
+class CameraDec(nn.Module):
+    """Decode the final cam token into a 9-D pose encoding.
+
+    Output layout: ``[T(3), quat_xyzw(4), fov_h, fov_w]``. The translation is
+    always predicted by the network; the quaternion and FoV can either be
+    predicted or supplied via ``camera_encoding`` (used at training time
+    when GT cameras are available -- not exercised at inference here).
+
+    Parameters mirror the upstream ``cam_dec.py`` so HF weights load directly.
+    """
+
+    def __init__(self, dim_in: int = 1536,
+                 *, device=None, dtype=None, operations=None, **_kwargs):
+        super().__init__()
+        d = dim_in
+        self.backbone = nn.Sequential(
+            operations.Linear(d, d, device=device, dtype=dtype),
+            nn.ReLU(),
+            operations.Linear(d, d, device=device, dtype=dtype),
+            nn.ReLU(),
+        )
+        self.fc_t = operations.Linear(d, 3, device=device, dtype=dtype)
+        self.fc_qvec = operations.Linear(d, 4, device=device, dtype=dtype)
+        self.fc_fov = nn.Sequential(
+            operations.Linear(d, 2, device=device, dtype=dtype),
+            nn.ReLU(),
+        )
+
+    def forward(self, feat: torch.Tensor,
+                camera_encoding: "torch.Tensor | None" = None) -> torch.Tensor:
+        """Decode ``(B, N, dim_in)`` cam tokens into ``(B, N, 9)`` pose enc."""
+        B, N = feat.shape[:2]
+        feat = feat.reshape(B * N, -1)
+        feat = self.backbone(feat)
+        out_t = self.fc_t(feat.float()).reshape(B, N, 3)
+        if camera_encoding is None:
+            out_qvec = self.fc_qvec(feat.float()).reshape(B, N, 4)
+            out_fov = self.fc_fov(feat.float()).reshape(B, N, 2)
+        else:
+            out_qvec = camera_encoding[..., 3:7]
+            out_fov = camera_encoding[..., -2:]
+        return torch.cat([out_t, out_qvec, out_fov], dim=-1)

comfy/ldm/depth_anything_3/dpt.py

@@ -0,0 +1,489 @@
+"""DPT / DualDPT heads for Depth Anything 3."""
+
+from __future__ import annotations
+
+from typing import List, Optional, Sequence, Tuple
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+
+class Permute(nn.Module):
+    def __init__(self, dims: Tuple[int, ...]):
+        super().__init__()
+        self.dims = dims
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        return x.permute(*self.dims)
+
+
+def _custom_interpolate(
+    x: torch.Tensor,
+    size: Optional[Tuple[int, int]] = None,
+    scale_factor: Optional[float] = None,
+    mode: str = "bilinear",
+    align_corners: bool = True,
+) -> torch.Tensor:
+    if size is None:
+        assert scale_factor is not None
+        size = (int(x.shape[-2] * scale_factor), int(x.shape[-1] * scale_factor))
+    INT_MAX = 1610612736
+    total = size[0] * size[1] * x.shape[0] * x.shape[1]
+    if total > INT_MAX:
+        chunks = torch.chunk(x, chunks=(total // INT_MAX) + 1, dim=0)
+        outs = [F.interpolate(c, size=size, mode=mode, align_corners=align_corners) for c in chunks]
+        return torch.cat(outs, dim=0).contiguous()
+    return F.interpolate(x, size=size, mode=mode, align_corners=align_corners)
+
+
+def _create_uv_grid(width: int, height: int, aspect_ratio: float, dtype, device) -> torch.Tensor:
+    """Normalised UV grid spanning (-x_span, -y_span)..(x_span, y_span)."""
+    diag_factor = (aspect_ratio ** 2 + 1.0) ** 0.5
+    span_x = aspect_ratio / diag_factor
+    span_y = 1.0 / diag_factor
+    left_x = -span_x * (width - 1) / width
+    right_x = span_x * (width - 1) / width
+    top_y = -span_y * (height - 1) / height
+    bottom_y = span_y * (height - 1) / height
+    x_coords = torch.linspace(left_x, right_x, steps=width, dtype=dtype, device=device)
+    y_coords = torch.linspace(top_y, bottom_y, steps=height, dtype=dtype, device=device)
+    uu, vv = torch.meshgrid(x_coords, y_coords, indexing="xy")
+    return torch.stack((uu, vv), dim=-1)  # (H, W, 2)
+
+
+def _make_sincos_pos_embed(embed_dim: int, pos: torch.Tensor, omega_0: float = 100.0) -> torch.Tensor:
+    omega = torch.arange(embed_dim // 2, dtype=torch.float32, device=pos.device)
+    omega = 1.0 / omega_0 ** (omega / (embed_dim / 2.0))
+    pos = pos.reshape(-1)
+    out = torch.einsum("m,d->md", pos, omega)
+    return torch.cat([out.sin(), out.cos()], dim=1).float()
+
+
+def _position_grid_to_embed(pos_grid: torch.Tensor, embed_dim: int, omega_0: float = 100.0) -> torch.Tensor:
+    H, W, _ = pos_grid.shape
+    pos_flat = pos_grid.reshape(-1, 2)
+    emb_x = _make_sincos_pos_embed(embed_dim // 2, pos_flat[:, 0], omega_0=omega_0)
+    emb_y = _make_sincos_pos_embed(embed_dim // 2, pos_flat[:, 1], omega_0=omega_0)
+    emb = torch.cat([emb_x, emb_y], dim=-1)
+    return emb.view(H, W, embed_dim)
+
+
+def _add_pos_embed(x: torch.Tensor, W: int, H: int, ratio: float = 0.1) -> torch.Tensor:
+    """Stateless UV positional embedding added to a feature map (B, C, h, w)."""
+    pw, ph = x.shape[-1], x.shape[-2]
+    pe = _create_uv_grid(pw, ph, aspect_ratio=W / H, dtype=x.dtype, device=x.device)
+    pe = _position_grid_to_embed(pe, x.shape[1]) * ratio
+    pe = pe.permute(2, 0, 1)[None].expand(x.shape[0], -1, -1, -1).to(dtype=x.dtype)
+    return x + pe
+
+
+def _apply_activation(x: torch.Tensor, activation: str) -> torch.Tensor:
+    act = (activation or "linear").lower()
+    if act == "exp":
+        return torch.exp(x)
+    if act == "expp1":
+        return torch.exp(x) + 1
+    if act == "expm1":
+        return torch.expm1(x)
+    if act == "relu":
+        return torch.relu(x)
+    if act == "sigmoid":
+        return torch.sigmoid(x)
+    if act == "softplus":
+        return F.softplus(x)
+    if act == "tanh":
+        return torch.tanh(x)
+    return x
+
+
+# -----------------------------------------------------------------------------
+# Fusion building blocks
+# -----------------------------------------------------------------------------
+
+
+class ResidualConvUnit(nn.Module):
+    def __init__(self, features: int, device=None, dtype=None, operations=None):
+        super().__init__()
+        self.conv1 = operations.Conv2d(features, features, 3, 1, 1, bias=True, device=device, dtype=dtype)
+        self.conv2 = operations.Conv2d(features, features, 3, 1, 1, bias=True, device=device, dtype=dtype)
+        self.activation = nn.ReLU(inplace=False)
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        out = self.activation(x)
+        out = self.conv1(out)
+        out = self.activation(out)
+        out = self.conv2(out)
+        return out + x
+
+
+class FeatureFusionBlock(nn.Module):
+    def __init__(self, features: int, has_residual: bool = True, align_corners: bool = True, device=None, dtype=None, operations=None):
+        super().__init__()
+        self.align_corners = align_corners
+        self.has_residual = has_residual
+        if has_residual:
+            self.resConfUnit1 = ResidualConvUnit(features, device=device, dtype=dtype, operations=operations)
+        else:
+            self.resConfUnit1 = None
+        self.resConfUnit2 = ResidualConvUnit(features, device=device, dtype=dtype, operations=operations)
+        self.out_conv = operations.Conv2d(features, features, 1, 1, 0, bias=True, device=device, dtype=dtype)
+
+    def forward(self, *xs: torch.Tensor, size: Optional[Tuple[int, int]] = None) -> torch.Tensor:
+        y = xs[0]
+        if self.has_residual and len(xs) > 1 and self.resConfUnit1 is not None:
+            y = y + self.resConfUnit1(xs[1])
+        y = self.resConfUnit2(y)
+        if size is None:
+            up_kwargs = {"scale_factor": 2.0}
+        else:
+            up_kwargs = {"size": size}
+        y = _custom_interpolate(y, **up_kwargs, mode="bilinear", align_corners=self.align_corners)
+        y = self.out_conv(y)
+        return y
+
+
+class _Scratch(nn.Module):
+    """Container that mirrors upstream ``scratch`` attribute layout."""
+
+
+def _make_scratch(in_shape: List[int], out_shape: int, device=None, dtype=None, operations=None) -> _Scratch:
+    scratch = _Scratch()
+    scratch.layer1_rn = operations.Conv2d(in_shape[0], out_shape, 3, 1, 1, bias=False, device=device, dtype=dtype)
+    scratch.layer2_rn = operations.Conv2d(in_shape[1], out_shape, 3, 1, 1, bias=False, device=device, dtype=dtype)
+    scratch.layer3_rn = operations.Conv2d(in_shape[2], out_shape, 3, 1, 1, bias=False, device=device, dtype=dtype)
+    scratch.layer4_rn = operations.Conv2d(in_shape[3], out_shape, 3, 1, 1, bias=False, device=device, dtype=dtype)
+    return scratch
+
+
+def _make_fusion_block(features: int, has_residual: bool = True, device=None, dtype=None, operations=None) -> FeatureFusionBlock:
+    return FeatureFusionBlock(features, has_residual=has_residual, align_corners=True, device=device, dtype=dtype, operations=operations)
+
+
+# -----------------------------------------------------------------------------
+# DPT (single head + optional sky head) -- used by DA3Mono/Metric
+# -----------------------------------------------------------------------------
+
+
+class DPT(nn.Module):
+    """Single-head DPT used by DA3Mono-Large and DA3Metric-Large."""
+
+    def __init__(
+        self,
+        dim_in: int,
+        patch_size: int = 14,
+        output_dim: int = 1,
+        activation: str = "exp",
+        conf_activation: str = "expp1",
+        features: int = 256,
+        out_channels: Sequence[int] = (256, 512, 1024, 1024),
+        pos_embed: bool = False,
+        down_ratio: int = 1,
+        head_name: str = "depth",
+        use_sky_head: bool = True,
+        sky_name: str = "sky",
+        sky_activation: str = "relu",
+        norm_type: str = "idt",
+        device=None, dtype=None, operations=None,
+    ):
+        super().__init__()
+        self.patch_size = patch_size
+        self.activation = activation
+        self.conf_activation = conf_activation
+        self.pos_embed = pos_embed
+        self.down_ratio = down_ratio
+        self.head_main = head_name
+        self.sky_name = sky_name
+        self.out_dim = output_dim
+        self.has_conf = output_dim > 1
+        self.use_sky_head = use_sky_head
+        self.sky_activation = sky_activation
+        self.intermediate_layer_idx: Tuple[int, int, int, int] = (0, 1, 2, 3)
+
+        if norm_type == "layer":
+            self.norm = operations.LayerNorm(dim_in, device=device, dtype=dtype)
+        else:
+            self.norm = nn.Identity()
+
+        out_channels = list(out_channels)
+        self.projects = nn.ModuleList([
+            operations.Conv2d(dim_in, oc, kernel_size=1, stride=1, padding=0, device=device, dtype=dtype)
+            for oc in out_channels
+        ])
+        self.resize_layers = nn.ModuleList([
+            operations.ConvTranspose2d(out_channels[0], out_channels[0], kernel_size=4, stride=4, padding=0, device=device, dtype=dtype),
+            operations.ConvTranspose2d(out_channels[1], out_channels[1], kernel_size=2, stride=2, padding=0, device=device, dtype=dtype),
+            nn.Identity(),
+            operations.Conv2d(out_channels[3], out_channels[3], kernel_size=3, stride=2, padding=1, device=device, dtype=dtype),
+        ])
+
+        self.scratch = _make_scratch(out_channels, features, device=device, dtype=dtype, operations=operations)
+        self.scratch.refinenet1 = _make_fusion_block(features, device=device, dtype=dtype, operations=operations)
+        self.scratch.refinenet2 = _make_fusion_block(features, device=device, dtype=dtype, operations=operations)
+        self.scratch.refinenet3 = _make_fusion_block(features, device=device, dtype=dtype, operations=operations)
+        self.scratch.refinenet4 = _make_fusion_block(features, has_residual=False, device=device, dtype=dtype, operations=operations)
+
+        head_features_1 = features
+        head_features_2 = 32
+        self.scratch.output_conv1 = operations.Conv2d(
+            head_features_1, head_features_1 // 2, kernel_size=3, stride=1, padding=1,
+            device=device, dtype=dtype,
+        )
+        self.scratch.output_conv2 = nn.Sequential(
+            operations.Conv2d(head_features_1 // 2, head_features_2, kernel_size=3, stride=1, padding=1, device=device, dtype=dtype),
+            nn.ReLU(inplace=False),
+            operations.Conv2d(head_features_2, output_dim, kernel_size=1, stride=1, padding=0, device=device, dtype=dtype),
+        )
+
+        if self.use_sky_head:
+            self.scratch.sky_output_conv2 = nn.Sequential(
+                operations.Conv2d(head_features_1 // 2, head_features_2, kernel_size=3, stride=1, padding=1, device=device, dtype=dtype),
+                nn.ReLU(inplace=False),
+                operations.Conv2d(head_features_2, 1, kernel_size=1, stride=1, padding=0, device=device, dtype=dtype),
+            )
+
+    def forward(self, feats: List[torch.Tensor], H: int, W: int, patch_start_idx: int = 0, **_kwargs) -> dict:
+        # feats[i][0] is the patch-token tensor with shape (B, S, N_patch, C)
+        B, S, N, C = feats[0][0].shape
+        feats_flat = [feat[0].reshape(B * S, N, C) for feat in feats]
+
+        ph, pw = H // self.patch_size, W // self.patch_size
+        resized = []
+        for stage_idx, take_idx in enumerate(self.intermediate_layer_idx):
+            x = feats_flat[take_idx][:, patch_start_idx:]
+            x = self.norm(x)
+            x = x.permute(0, 2, 1).contiguous().reshape(B * S, C, ph, pw)
+            x = self.projects[stage_idx](x)
+            if self.pos_embed:
+                x = _add_pos_embed(x, W, H)
+            x = self.resize_layers[stage_idx](x)
+            resized.append(x)
+
+        l1_rn = self.scratch.layer1_rn(resized[0])
+        l2_rn = self.scratch.layer2_rn(resized[1])
+        l3_rn = self.scratch.layer3_rn(resized[2])
+        l4_rn = self.scratch.layer4_rn(resized[3])
+
+        out = self.scratch.refinenet4(l4_rn, size=l3_rn.shape[2:])
+        out = self.scratch.refinenet3(out, l3_rn, size=l2_rn.shape[2:])
+        out = self.scratch.refinenet2(out, l2_rn, size=l1_rn.shape[2:])
+        out = self.scratch.refinenet1(out, l1_rn)
+
+        h_out = int(ph * self.patch_size / self.down_ratio)
+        w_out = int(pw * self.patch_size / self.down_ratio)
+
+        fused = self.scratch.output_conv1(out)
+        fused = _custom_interpolate(fused, (h_out, w_out), mode="bilinear", align_corners=True)
+        if self.pos_embed:
+            fused = _add_pos_embed(fused, W, H)
+        feat = fused
+
+        main_logits = self.scratch.output_conv2(feat)
+        outs = {}
+        if self.has_conf:
+            fmap = main_logits.permute(0, 2, 3, 1)
+            pred = _apply_activation(fmap[..., :-1], self.activation)
+            conf = _apply_activation(fmap[..., -1], self.conf_activation)
+            outs[self.head_main] = pred.squeeze(-1).view(B, S, *pred.shape[1:-1])
+            outs[f"{self.head_main}_conf"] = conf.view(B, S, *conf.shape[1:])
+        else:
+            pred = _apply_activation(main_logits, self.activation)
+            outs[self.head_main] = pred.squeeze(1).view(B, S, *pred.shape[2:])
+
+        if self.use_sky_head:
+            sky_logits = self.scratch.sky_output_conv2(feat)
+            if self.sky_activation.lower() == "sigmoid":
+                sky = torch.sigmoid(sky_logits)
+            elif self.sky_activation.lower() == "relu":
+                sky = F.relu(sky_logits)
+            else:
+                sky = sky_logits
+            outs[self.sky_name] = sky.squeeze(1).view(B, S, *sky.shape[2:])
+
+        return outs
+
+
+# -----------------------------------------------------------------------------
+# DualDPT (depth + auxiliary "ray" head) -- used by DA3-Small / DA3-Base
+# -----------------------------------------------------------------------------
+
+
+class DualDPT(nn.Module):
+    """Two-head DPT used by DA3-Small / DA3-Base."""
+
+    def __init__(
+        self,
+        dim_in: int,
+        patch_size: int = 14,
+        output_dim: int = 2,
+        activation: str = "exp",
+        conf_activation: str = "expp1",
+        features: int = 256,
+        out_channels: Sequence[int] = (256, 512, 1024, 1024),
+        pos_embed: bool = True,
+        down_ratio: int = 1,
+        aux_pyramid_levels: int = 4,
+        aux_out1_conv_num: int = 5,
+        head_names: Tuple[str, str] = ("depth", "ray"),
+        device=None, dtype=None, operations=None,
+    ):
+        super().__init__()
+        self.patch_size = patch_size
+        self.activation = activation
+        self.conf_activation = conf_activation
+        self.pos_embed = pos_embed
+        self.down_ratio = down_ratio
+        self.aux_levels = aux_pyramid_levels
+        self.aux_out1_conv_num = aux_out1_conv_num
+        self.head_main, self.head_aux = head_names
+        self.intermediate_layer_idx: Tuple[int, int, int, int] = (0, 1, 2, 3)
+        # Toggle the auxiliary ray branch at runtime. Default off (mono path).
+        # DepthAnything3Net flips this on when running multi-view + ray-pose.
+        self.enable_aux: bool = False
+
+        self.norm = operations.LayerNorm(dim_in, device=device, dtype=dtype)
+        out_channels = list(out_channels)
+        self.projects = nn.ModuleList([
+            operations.Conv2d(dim_in, oc, kernel_size=1, stride=1, padding=0, device=device, dtype=dtype)
+            for oc in out_channels
+        ])
+        self.resize_layers = nn.ModuleList([
+            operations.ConvTranspose2d(out_channels[0], out_channels[0], kernel_size=4, stride=4, padding=0, device=device, dtype=dtype),
+            operations.ConvTranspose2d(out_channels[1], out_channels[1], kernel_size=2, stride=2, padding=0, device=device, dtype=dtype),
+            nn.Identity(),
+            operations.Conv2d(out_channels[3], out_channels[3], kernel_size=3, stride=2, padding=1, device=device, dtype=dtype),
+        ])
+
+        self.scratch = _make_scratch(out_channels, features, device=device, dtype=dtype, operations=operations)
+        # Main fusion chain
+        self.scratch.refinenet1 = _make_fusion_block(features, device=device, dtype=dtype, operations=operations)
+        self.scratch.refinenet2 = _make_fusion_block(features, device=device, dtype=dtype, operations=operations)
+        self.scratch.refinenet3 = _make_fusion_block(features, device=device, dtype=dtype, operations=operations)
+        self.scratch.refinenet4 = _make_fusion_block(features, has_residual=False, device=device, dtype=dtype, operations=operations)
+        # Auxiliary fusion chain (separate copies)
+        self.scratch.refinenet1_aux = _make_fusion_block(features, device=device, dtype=dtype, operations=operations)
+        self.scratch.refinenet2_aux = _make_fusion_block(features, device=device, dtype=dtype, operations=operations)
+        self.scratch.refinenet3_aux = _make_fusion_block(features, device=device, dtype=dtype, operations=operations)
+        self.scratch.refinenet4_aux = _make_fusion_block(features, has_residual=False, device=device, dtype=dtype, operations=operations)
+
+        head_features_1 = features
+        head_features_2 = 32
+
+        # Main head neck + final projection
+        self.scratch.output_conv1 = operations.Conv2d(
+            head_features_1, head_features_1 // 2, kernel_size=3, stride=1, padding=1,
+            device=device, dtype=dtype,
+        )
+        self.scratch.output_conv2 = nn.Sequential(
+            operations.Conv2d(head_features_1 // 2, head_features_2, kernel_size=3, stride=1, padding=1, device=device, dtype=dtype),
+            nn.ReLU(inplace=False),
+            operations.Conv2d(head_features_2, output_dim, kernel_size=1, stride=1, padding=0, device=device, dtype=dtype),
+        )
+
+        # Aux pre-head per level (multi-level pyramid)
+        self.scratch.output_conv1_aux = nn.ModuleList([
+            self._make_aux_out1_block(head_features_1, device=device, dtype=dtype, operations=operations)
+            for _ in range(self.aux_levels)
+        ])
+
+        # Aux final projection per level (includes LayerNorm permute path).
+        ln_seq = [Permute((0, 2, 3, 1)),
+                  operations.LayerNorm(head_features_2, device=device, dtype=dtype),
+                  Permute((0, 3, 1, 2))]
+        self.scratch.output_conv2_aux = nn.ModuleList([
+            nn.Sequential(
+                operations.Conv2d(head_features_1 // 2, head_features_2, kernel_size=3, stride=1, padding=1, device=device, dtype=dtype),
+                *ln_seq,
+                nn.ReLU(inplace=False),
+                operations.Conv2d(head_features_2, 7, kernel_size=1, stride=1, padding=0, device=device, dtype=dtype),
+            )
+            for _ in range(self.aux_levels)
+        ])
+
+    @staticmethod
+    def _make_aux_out1_block(in_ch: int, *, device=None, dtype=None, operations=None) -> nn.Sequential:
+        # aux_out1_conv_num=5 in all Apache-2.0 variants.
+        return nn.Sequential(
+            operations.Conv2d(in_ch, in_ch // 2, 3, 1, 1, device=device, dtype=dtype),
+            operations.Conv2d(in_ch // 2, in_ch, 3, 1, 1, device=device, dtype=dtype),
+            operations.Conv2d(in_ch, in_ch // 2, 3, 1, 1, device=device, dtype=dtype),
+            operations.Conv2d(in_ch // 2, in_ch, 3, 1, 1, device=device, dtype=dtype),
+            operations.Conv2d(in_ch, in_ch // 2, 3, 1, 1, device=device, dtype=dtype),
+        )
+
+    def forward(self, feats: List[torch.Tensor], H: int, W: int, patch_start_idx: int = 0, **_kwargs) -> dict:
+        B, S, N, C = feats[0][0].shape
+        feats_flat = [feat[0].reshape(B * S, N, C) for feat in feats]
+
+        ph, pw = H // self.patch_size, W // self.patch_size
+        resized = []
+        for stage_idx, take_idx in enumerate(self.intermediate_layer_idx):
+            x = feats_flat[take_idx][:, patch_start_idx:]
+            x = self.norm(x)
+            x = x.permute(0, 2, 1).contiguous().reshape(B * S, C, ph, pw)
+            x = self.projects[stage_idx](x)
+            if self.pos_embed:
+                x = _add_pos_embed(x, W, H)
+            x = self.resize_layers[stage_idx](x)
+            resized.append(x)
+
+        l1_rn = self.scratch.layer1_rn(resized[0])
+        l2_rn = self.scratch.layer2_rn(resized[1])
+        l3_rn = self.scratch.layer3_rn(resized[2])
+        l4_rn = self.scratch.layer4_rn(resized[3])
+
+        # Main pyramid (output_conv1 is applied inside the upstream `_fuse`,
+        # before interpolation -- replicate that order here).
+        m = self.scratch.refinenet4(l4_rn, size=l3_rn.shape[2:])
+        if self.enable_aux:
+            a4 = self.scratch.refinenet4_aux(l4_rn, size=l3_rn.shape[2:])
+            aux_pyr = [a4]
+        m = self.scratch.refinenet3(m, l3_rn, size=l2_rn.shape[2:])
+        if self.enable_aux:
+            aux_pyr.append(self.scratch.refinenet3_aux(aux_pyr[-1], l3_rn, size=l2_rn.shape[2:]))
+        m = self.scratch.refinenet2(m, l2_rn, size=l1_rn.shape[2:])
+        if self.enable_aux:
+            aux_pyr.append(self.scratch.refinenet2_aux(aux_pyr[-1], l2_rn, size=l1_rn.shape[2:]))
+        m = self.scratch.refinenet1(m, l1_rn)
+        if self.enable_aux:
+            aux_pyr.append(self.scratch.refinenet1_aux(aux_pyr[-1], l1_rn))
+        m = self.scratch.output_conv1(m)
+
+        h_out = int(ph * self.patch_size / self.down_ratio)
+        w_out = int(pw * self.patch_size / self.down_ratio)
+
+        m = _custom_interpolate(m, (h_out, w_out), mode="bilinear", align_corners=True)
+        if self.pos_embed:
+            m = _add_pos_embed(m, W, H)
+        main_logits = self.scratch.output_conv2(m)
+        fmap = main_logits.permute(0, 2, 3, 1)
+        depth_pred = _apply_activation(fmap[..., :-1], self.activation)
+        depth_conf = _apply_activation(fmap[..., -1], self.conf_activation)
+
+        outs = {
+            self.head_main: depth_pred.squeeze(-1).view(B, S, *depth_pred.shape[1:-1]),
+            f"{self.head_main}_conf": depth_conf.view(B, S, *depth_conf.shape[1:]),
+        }
+
+        if self.enable_aux:
+            # Auxiliary "ray" head (multi-level inside) -- only the last level
+            # is returned. Mirrors upstream ``DualDPT._fuse`` + ``_forward_impl``:
+            # each aux pyramid level goes through ``output_conv1_aux[i]``
+            # (5-layer conv stack that ends at ``features // 2`` channels),
+            # then the last level optionally gets a pos-embed and finally
+            # ``output_conv2_aux[-1]``.
+            aux_processed = [
+                self.scratch.output_conv1_aux[i](a) for i, a in enumerate(aux_pyr)
+            ]
+            last_aux = aux_processed[-1]
+            if self.pos_embed:
+                last_aux = _add_pos_embed(last_aux, W, H)
+            last_aux_logits = self.scratch.output_conv2_aux[-1](last_aux)
+            fmap_last = last_aux_logits.permute(0, 2, 3, 1)
+            # Channels: [ray(6), ray_conf(1)]; ray uses 'linear' activation.
+            aux_pred = fmap_last[..., :-1]
+            aux_conf = _apply_activation(fmap_last[..., -1], self.conf_activation)
+            outs[self.head_aux] = aux_pred.view(B, S, *aux_pred.shape[1:])
+            outs[f"{self.head_aux}_conf"] = aux_conf.view(B, S, *aux_conf.shape[1:])
+
+        return outs

comfy/ldm/depth_anything_3/model.py

@@ -0,0 +1,236 @@
+from __future__ import annotations
+
+from typing import Dict, Optional, Sequence
+
+import torch
+import torch.nn as nn
+
+from comfy.image_encoders.dino2 import Dinov2Model
+
+from .camera import CameraDec, CameraEnc
+from .dpt import DPT, DualDPT
+from .ray_pose import get_extrinsic_from_camray
+from .transform import affine_inverse, pose_encoding_to_extri_intri
+
+
+_HEAD_REGISTRY = {
+    "dpt": DPT,
+    "dualdpt": DualDPT,
+}
+
+
+# Backbone presets (mirror the upstream DINOv2 ViT variants).
+_BACKBONE_PRESETS = {
+    "vits": dict(hidden_size=384,  num_hidden_layers=12, num_attention_heads=6,  use_swiglu_ffn=False),
+    "vitb": dict(hidden_size=768,  num_hidden_layers=12, num_attention_heads=12, use_swiglu_ffn=False),
+    "vitl": dict(hidden_size=1024, num_hidden_layers=24, num_attention_heads=16, use_swiglu_ffn=False),
+    "vitg": dict(hidden_size=1536, num_hidden_layers=40, num_attention_heads=24, use_swiglu_ffn=True),
+}
+
+
+def _build_backbone_config(
+    backbone_name: str,
+    *,
+    alt_start: int,
+    qknorm_start: int,
+    rope_start: int,
+    cat_token: bool,
+) -> dict:
+    if backbone_name not in _BACKBONE_PRESETS:
+        raise ValueError(f"Unknown DINOv2 backbone variant: {backbone_name!r}")
+    cfg = dict(_BACKBONE_PRESETS[backbone_name])
+    cfg.update(dict(
+        layer_norm_eps=1e-6,
+        patch_size=14,
+        image_size=518,
+        # No mask_token in DA3 weights; omit param to avoid load warnings.
+        use_mask_token=False,
+        alt_start=alt_start,
+        qknorm_start=qknorm_start,
+        rope_start=rope_start,
+        cat_token=cat_token,
+        rope_freq=100.0,
+    ))
+    return cfg
+
+
+class DepthAnything3Net(nn.Module):
+
+    PATCH_SIZE = 14
+
+    def __init__(
+        self,
+        # --- Backbone ---
+        backbone_name: str = "vitl",
+        out_layers: Sequence[int] = (4, 11, 17, 23),
+        alt_start: int = -1,
+        qknorm_start: int = -1,
+        rope_start: int = -1,
+        cat_token: bool = False,
+        # --- Head ---
+        head_type: str = "dpt",  # dpt or dualdpt
+        head_dim_in: int = 1024,
+        head_output_dim: int = 1,  # 1 = depth only, 2 = depth+conf
+        head_features: int = 256,
+        head_out_channels: Sequence[int] = (256, 512, 1024, 1024),
+        head_use_sky_head: bool = True, # ignored by DualDPT
+        head_pos_embed: Optional[bool] = None,  # default: True for DualDPT, False for DPT
+        # --- Camera (multi-view) ---
+        has_cam_enc: bool = False,
+        has_cam_dec: bool = False,
+        cam_dim_out: Optional[int] = None,  # CameraEnc dim_out (defaults to embed_dim)
+        cam_dec_dim_in: Optional[int] = None,  # CameraDec dim_in  (defaults to 2*embed_dim with cat_token)
+        # ComfyUI plumbing
+        device=None, dtype=None, operations=None,
+        **_ignored,
+    ):
+        super().__init__()
+        head_cls = _HEAD_REGISTRY[head_type.lower()]
+        self.head_type = head_type.lower()
+        self.has_sky = (self.head_type == "dpt") and head_use_sky_head
+        self.has_conf = head_output_dim > 1
+        self.out_layers = list(out_layers)
+
+        backbone_cfg = _build_backbone_config(
+            backbone_name,
+            alt_start=alt_start,
+            qknorm_start=qknorm_start,
+            rope_start=rope_start,
+            cat_token=cat_token,
+        )
+        self.backbone = Dinov2Model(backbone_cfg, dtype, device, operations)
+
+        head_kwargs = dict(
+            dim_in=head_dim_in,
+            patch_size=self.PATCH_SIZE,
+            output_dim=head_output_dim,
+            features=head_features,
+            out_channels=tuple(head_out_channels),
+            device=device, dtype=dtype, operations=operations,
+        )
+        if self.head_type == "dpt":
+            head_kwargs.update(
+                use_sky_head=head_use_sky_head,
+                pos_embed=(False if head_pos_embed is None else head_pos_embed),
+            )
+        else:  # dualdpt
+            head_kwargs.update(
+                pos_embed=(True if head_pos_embed is None else head_pos_embed),
+            )
+        self.head = head_cls(**head_kwargs)
+
+        # Built only if checkpoint has weights; cam_enc output dim == embed_dim.
+        embed_dim = backbone_cfg["hidden_size"]
+        if has_cam_enc:
+            self.cam_enc = CameraEnc(
+                dim_out=cam_dim_out if cam_dim_out is not None else embed_dim,
+                num_heads=max(1, embed_dim // 64),
+                device=device, dtype=dtype, operations=operations,
+            )
+        else:
+            self.cam_enc = None
+        if has_cam_dec:
+            default_dim = embed_dim * (2 if cat_token else 1)
+            self.cam_dec = CameraDec(
+                dim_in=cam_dec_dim_in if cam_dec_dim_in is not None else default_dim,
+                device=device, dtype=dtype, operations=operations,
+            )
+        else:
+            self.cam_dec = None
+
+        self.dtype = dtype
+
+    def forward(
+        self,
+        image: torch.Tensor,
+        extrinsics: Optional[torch.Tensor] = None,
+        intrinsics: Optional[torch.Tensor] = None,
+        *,
+        use_ray_pose: bool = False,
+        ref_view_strategy: str = "saddle_balanced",
+        export_feat_layers: Optional[Sequence[int]] = None,
+        **_unused,
+    ) -> Dict[str, torch.Tensor]:
+        """Run depth and optionally pose prediction."""
+        if image.ndim == 4:
+            image = image.unsqueeze(1)  # (B, 1, 3, H, W)
+        assert image.ndim == 5 and image.shape[2] == 3, \
+            f"image must be (B,3,H,W) or (B,S,3,H,W); got {tuple(image.shape)}"
+
+        B, S, _, H, W = image.shape
+        assert H % self.PATCH_SIZE == 0 and W % self.PATCH_SIZE == 0, \
+            f"image H,W must be multiples of {self.PATCH_SIZE}; got {(H, W)}"
+
+        # Camera-token preparation (multi-view path).
+        cam_token = None
+        if extrinsics is not None and intrinsics is not None and self.cam_enc is not None:
+            cam_token = self.cam_enc(extrinsics, intrinsics, (H, W))
+
+        # Toggle aux ray output on/off depending on what the caller asked for.
+        if isinstance(self.head, DualDPT):
+            self.head.enable_aux = bool(use_ray_pose)
+
+        feats, aux_feats = self.backbone.get_intermediate_layers_da3(
+            image, self.out_layers, cam_token=cam_token,
+            ref_view_strategy=ref_view_strategy,
+            export_feat_layers=export_feat_layers,
+        )
+        head_out = self.head(feats, H=H, W=W, patch_start_idx=0)
+
+        # Pose prediction.
+        out: Dict[str, torch.Tensor] = {}
+        if use_ray_pose and "ray" in head_out and "ray_conf" in head_out:
+            ray = head_out["ray"]
+            ray_conf = head_out["ray_conf"]
+            extr_c2w, focal, pp = get_extrinsic_from_camray(
+                ray, ray_conf, ray.shape[-3], ray.shape[-2],
+            )
+            # Match the upstream output: w2c, drop the homogeneous row.
+            extr_w2c = affine_inverse(extr_c2w)[:, :, :3, :]
+            # Build pixel-space intrinsics from the normalised focal/pp output.
+            intr = torch.eye(3, device=ray.device, dtype=ray.dtype)
+            intr = intr[None, None].expand(extr_c2w.shape[0], extr_c2w.shape[1], 3, 3).clone()
+            intr[:, :, 0, 0] = focal[:, :, 0] / 2 * W
+            intr[:, :, 1, 1] = focal[:, :, 1] / 2 * H
+            intr[:, :, 0, 2] = pp[:, :, 0] * W * 0.5
+            intr[:, :, 1, 2] = pp[:, :, 1] * H * 0.5
+            out["extrinsics"] = extr_w2c
+            out["intrinsics"] = intr
+        elif self.cam_dec is not None and S > 1:
+            # Decode the cam-token of the final out_layer into a pose encoding.
+            cam_feat = feats[-1][1]  # (B, S, dim_in_to_cam_dec)
+            pose_enc = self.cam_dec(cam_feat)
+            c2w_3x4, intr = pose_encoding_to_extri_intri(pose_enc, (H, W))
+            # Match the upstream output convention: w2c (world->camera), 3x4.
+            c2w_4x4 = torch.cat([
+                c2w_3x4,
+                torch.tensor([0, 0, 0, 1], device=c2w_3x4.device, dtype=c2w_3x4.dtype)
+                    .view(1, 1, 1, 4).expand(B, S, 1, 4),
+            ], dim=-2)
+            out["extrinsics"] = affine_inverse(c2w_4x4)[:, :, :3, :]
+            out["intrinsics"] = intr
+
+        # Flatten the views axis for per-pixel outputs (depth/conf/sky) so the
+        # per-image consumer keeps its (B*S, H, W) interface.
+        for k, v in head_out.items():
+            if k in ("ray", "ray_conf"):
+                # Keep multi-view shape for downstream pose work.
+                out[k] = v
+            elif v.ndim >= 3 and v.shape[0] == B and v.shape[1] == S:
+                out[k] = v.reshape(B * S, *v.shape[2:])
+            else:
+                out[k] = v
+
+        if export_feat_layers:
+            out["aux_features"] = self._reshape_aux_features(aux_feats, H, W)
+        return out
+
+    def _reshape_aux_features(self, aux_feats, H: int, W: int):
+        """Reshape (B, S, N, C) aux features into (B, S, h_p, w_p, C)."""
+        ph, pw = H // self.PATCH_SIZE, W // self.PATCH_SIZE
+        out = []
+        for f in aux_feats:
+            B, S, N, C = f.shape
+            assert N == ph * pw, f"aux feature seq mismatch: {N} != {ph}*{pw}"
+            out.append(f.reshape(B, S, ph, pw, C))
+        return out

comfy/ldm/depth_anything_3/preprocess.py

@@ -0,0 +1,128 @@
+"""Input/output preprocessing helpers for Depth Anything 3."""
+
+from __future__ import annotations
+
+from typing import Tuple
+
+import torch
+
+import comfy.utils
+
+PATCH_SIZE = 14
+
+# ImageNet normalization constants used during DA3 training.
+_IMAGENET_MEAN = torch.tensor([0.485, 0.456, 0.406])
+_IMAGENET_STD = torch.tensor([0.229, 0.224, 0.225])
+
+
+def _round_to_patch(x: int, patch: int = PATCH_SIZE) -> int:
+    down = (x // patch) * patch
+    up = down + patch
+    return up if abs(up - x) <= abs(x - down) else down
+
+
+def compute_target_size(orig_h: int, orig_w: int, process_res: int, method: str = "upper_bound_resize") -> Tuple[int, int]:
+    """Compute (target_h, target_w) for a single image.
+    upper_bound_resize: scale longest side to process_res, then round each dim to nearest multiple of 14 (default upstream method).
+    lower_bound_resize: scale shortest side to process_res, then round."""
+
+    if method == "upper_bound_resize":
+        longest = max(orig_h, orig_w)
+        scale = process_res / float(longest)
+    elif method == "lower_bound_resize":
+        shortest = min(orig_h, orig_w)
+        scale = process_res / float(shortest)
+    else:
+        raise ValueError(f"Unsupported process_res_method: {method}")
+
+    new_w = max(1, _round_to_patch(int(round(orig_w * scale))))
+    new_h = max(1, _round_to_patch(int(round(orig_h * scale))))
+    return new_h, new_w
+
+
+def preprocess_image(image: torch.Tensor, process_res: int = 504, method: str = "upper_bound_resize") -> torch.Tensor:
+    assert image.ndim == 4 and image.shape[-1] == 3, f"expected (B,H,W,3) IMAGE; got {tuple(image.shape)}"
+    B, H, W, _ = image.shape
+    target_h, target_w = compute_target_size(H, W, process_res, method)
+
+    # (B, H, W, 3) -> (B, 3, H, W)
+    x = image.movedim(-1, 1).contiguous()
+    if (target_h, target_w) != (H, W):
+        # Upstream uses cv2 INTER_CUBIC (upscale) / INTER_AREA (downscale).
+        # Lanczos in ``common_upscale`` is anti-aliased and produces the
+        # closest pixel-wise match in a sweep across {bilinear, bicubic,
+        # area, lanczos, bislerp}. Used in both directions for simplicity.
+        x = comfy.utils.common_upscale(x.float(), target_w, target_h, "lanczos", "disabled",)
+    x = x.clamp(0.0, 1.0)
+
+    mean = _IMAGENET_MEAN.to(device=x.device, dtype=x.dtype).view(1, 3, 1, 1)
+    std = _IMAGENET_STD.to(device=x.device, dtype=x.dtype).view(1, 3, 1, 1)
+    x = (x - mean) / std
+    return x
+
+
+# -----------------------------------------------------------------------------
+# Output post-processing (sky-aware clipping for Mono/Metric variants)
+# -----------------------------------------------------------------------------
+
+
+def compute_non_sky_mask(sky_prediction: torch.Tensor, threshold: float = 0.3) -> torch.Tensor:
+    """Boolean mask: True for non-sky pixels (sky probability < threshold)."""
+    return sky_prediction < threshold
+
+
+def apply_sky_aware_clip(depth: torch.Tensor, sky: torch.Tensor, threshold: float = 0.3, quantile: float = 0.99) -> torch.Tensor:
+    """Clips sky regions to the 99th percentile of non-sky depth. Returns a new depth tensor."""
+    non_sky = compute_non_sky_mask(sky, threshold=threshold)
+    if non_sky.sum() <= 10 or (~non_sky).sum() <= 10:
+        return depth.clone()
+
+    non_sky_depth = depth[non_sky]
+    if non_sky_depth.numel() > 100_000:
+        idx = torch.randint(0, non_sky_depth.numel(), (100_000,), device=non_sky_depth.device)
+        sampled = non_sky_depth[idx]
+    else:
+        sampled = non_sky_depth
+
+    max_depth = torch.quantile(sampled, quantile)
+    out = depth.clone()
+    out[~non_sky] = max_depth
+    return out
+
+
+def normalize_depth_v2_style(depth: torch.Tensor, sky: torch.Tensor | None = None, low_quantile: float = 0.01, high_quantile: float = 0.99) -> torch.Tensor:
+    """V2-style normalization computes percentile bounds over non-sky pixels (when available), then maps depth into [0, 1] with near = white (1.0)."""
+    if sky is not None:
+        mask = compute_non_sky_mask(sky)
+        if mask.any():
+            valid = depth[mask]
+        else:
+            valid = depth.flatten()
+    else:
+        valid = depth.flatten()
+
+    if valid.numel() > 100_000:
+        idx = torch.randint(0, valid.numel(), (100_000,), device=valid.device)
+        sample = valid[idx]
+    else:
+        sample = valid
+
+    lo = torch.quantile(sample, low_quantile)
+    hi = torch.quantile(sample, high_quantile)
+    rng = (hi - lo).clamp(min=1e-6)
+    norm = ((depth - lo) / rng).clamp(0.0, 1.0)
+    # Nearer pixels are brighter (1.0)
+    norm = 1.0 - norm
+    if sky is not None:
+        # Sky pixels become black (far / unknown)
+        sky_mask = ~compute_non_sky_mask(sky)
+        norm = torch.where(sky_mask, torch.zeros_like(norm), norm)
+    return norm
+
+
+def normalize_depth_min_max(depth: torch.Tensor) -> torch.Tensor:
+    """Simple per-frame min/max normalization with near=1.0 convention."""
+    lo = depth.amin(dim=(-2, -1), keepdim=True)
+    hi = depth.amax(dim=(-2, -1), keepdim=True)
+    rng = (hi - lo).clamp(min=1e-6)
+    return 1.0 - ((depth - lo) / rng).clamp(0.0, 1.0)

comfy/ldm/depth_anything_3/ray_pose.py

@@ -0,0 +1,272 @@
+"""Ray-to-pose conversion for the multi-view path of Depth Anything 3."""
+
+from __future__ import annotations
+
+from typing import Optional, Tuple
+
+import torch
+
+
+# qr/svd use fp32: CUDA often has no fp16/bf16 kernels for these ops.
+
+
+def _ql_decomposition(A: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor]:
+    """Decompose A = Q @ L with Q orthogonal and L lower-triangular.
+    Implemented in terms of QR by reversing the columns/rows; the standard
+    trick from the upstream reference. Inputs A are (3, 3)."""
+    P = torch.tensor([[0, 0, 1], [0, 1, 0], [1, 0, 0]], device=A.device, dtype=A.dtype)
+    A_tilde = A @ P
+    # CUDA QR is not implemented for fp16/bf16; upcast just for this call.
+    Q_tilde, R_tilde = torch.linalg.qr(A_tilde.float())
+    Q_tilde = Q_tilde.to(A.dtype)
+    R_tilde = R_tilde.to(A.dtype)
+    Q = Q_tilde @ P
+    L = P @ R_tilde @ P
+    d = torch.diag(L)
+    sign = torch.sign(d)
+    Q = Q * sign[None, :]  # scale columns of Q
+    L = L * sign[:, None]  # scale rows of L
+    return Q, L
+
+
+def _homogenize_points(points: torch.Tensor) -> torch.Tensor:
+    return torch.cat([points, torch.ones_like(points[..., :1])], dim=-1)
+
+
+# -----------------------------------------------------------------------------
+# Weighted-LSQ + RANSAC homography (batched)
+# -----------------------------------------------------------------------------
+
+
+def _find_homography_weighted_lsq(src_pts: torch.Tensor, dst_pts: torch.Tensor, confident_weight: torch.Tensor,) -> torch.Tensor:
+    """Solve a single H with weighted least-squares (DLT)."""
+    N = src_pts.shape[0]
+    if N < 4:
+        raise ValueError("At least 4 points are required to compute a homography.")
+    w = confident_weight.sqrt().unsqueeze(1)  # (N, 1)
+    x = src_pts[:, 0:1]
+    y = src_pts[:, 1:2]
+    u = dst_pts[:, 0:1]
+    v = dst_pts[:, 1:2]
+    zeros = torch.zeros_like(x)
+    A1 = torch.cat([-x * w, -y * w, -w, zeros, zeros, zeros, x * u * w, y * u * w, u * w], dim=1)
+    A2 = torch.cat([zeros, zeros, zeros, -x * w, -y * w, -w, x * v * w, y * v * w, v * w], dim=1)
+    A = torch.cat([A1, A2], dim=0)        # (2N, 9)
+    # CUDA SVD is not implemented for fp16/bf16; upcast just for this call.
+    _, _, Vh = torch.linalg.svd(A.float())
+    Vh = Vh.to(A.dtype)
+    H = Vh[-1].reshape(3, 3)
+    return H / H[-1, -1]
+
+
+def _find_homography_weighted_lsq_batched(src_pts_batch: torch.Tensor, dst_pts_batch: torch.Tensor, confident_weight_batch: torch.Tensor) -> torch.Tensor:
+    """Batched DLT solver. Inputs (B, K, 2) / (B, K); output (B, 3, 3)."""
+    B, K, _ = src_pts_batch.shape
+    w = confident_weight_batch.sqrt().unsqueeze(2)
+    x = src_pts_batch[:, :, 0:1]
+    y = src_pts_batch[:, :, 1:2]
+    u = dst_pts_batch[:, :, 0:1]
+    v = dst_pts_batch[:, :, 1:2]
+    zeros = torch.zeros_like(x)
+    A1 = torch.cat([-x * w, -y * w, -w, zeros, zeros, zeros, x * u * w, y * u * w, u * w], dim=2)
+    A2 = torch.cat([zeros, zeros, zeros, -x * w, -y * w, -w, x * v * w, y * v * w, v * w], dim=2)
+    A = torch.cat([A1, A2], dim=1)        # (B, 2K, 9)
+    # CUDA SVD is not implemented for fp16/bf16; upcast just for this call.
+    _, _, Vh = torch.linalg.svd(A.float())
+    Vh = Vh.to(A.dtype)
+    H = Vh[:, -1].reshape(B, 3, 3)
+    return H / H[:, 2:3, 2:3]
+
+
+def _ransac_find_homography_weighted_batched(
+    src_pts: torch.Tensor,                # (B, N, 2)
+    dst_pts: torch.Tensor,                # (B, N, 2)
+    confident_weight: torch.Tensor,       # (B, N)
+    n_sample: int,
+    n_iter: int = 100,
+    reproj_threshold: float = 3.0,
+    num_sample_for_ransac: int = 8,
+    max_inlier_num: int = 10000,
+    rand_sample_iters_idx: Optional[torch.Tensor] = None,
+) -> torch.Tensor:
+    """Batched weighted-RANSAC homography estimator. Returns (B, 3, 3) homography matrices."""
+    B, N, _ = src_pts.shape
+    assert N >= 4
+    device = src_pts.device
+
+    sorted_idx = torch.argsort(confident_weight, descending=True, dim=1)
+    candidate_idx = sorted_idx[:, :n_sample]                  # (B, n_sample)
+
+    if rand_sample_iters_idx is None:
+        rand_sample_iters_idx = torch.stack(
+            [torch.randperm(n_sample, device=device)[:num_sample_for_ransac]
+             for _ in range(n_iter)],
+            dim=0,
+        )
+
+    rand_idx = candidate_idx[:, rand_sample_iters_idx]        # (B, n_iter, k)
+    b_idx = (
+        torch.arange(B, device=device)
+        .view(B, 1, 1)
+        .expand(B, n_iter, num_sample_for_ransac)
+    )
+    src_b = src_pts[b_idx, rand_idx]
+    dst_b = dst_pts[b_idx, rand_idx]
+    w_b = confident_weight[b_idx, rand_idx]
+
+    cB, cN = src_b.shape[:2]
+    H_batch = _find_homography_weighted_lsq_batched(
+        src_b.flatten(0, 1), dst_b.flatten(0, 1), w_b.flatten(0, 1),
+    ).unflatten(0, (cB, cN))                                  # (B, n_iter, 3, 3)
+
+    src_homo = torch.cat([src_pts, torch.ones(B, N, 1, device=device, dtype=src_pts.dtype)], dim=2)
+    proj = torch.bmm(
+        src_homo.unsqueeze(1).expand(B, n_iter, N, 3).reshape(-1, N, 3),
+        H_batch.reshape(-1, 3, 3).transpose(1, 2),
+    )                                                          # (B*n_iter, N, 3)
+    proj_xy = (proj[:, :, :2] / proj[:, :, 2:3]).reshape(B, n_iter, N, 2)
+    err = ((proj_xy - dst_pts.unsqueeze(1)) ** 2).sum(-1).sqrt()  # (B, n_iter, N)
+    inlier_mask = err < reproj_threshold
+    score = (inlier_mask * confident_weight.unsqueeze(1)).sum(dim=2)
+    best_idx = torch.argmax(score, dim=1)
+    best_inlier_mask = inlier_mask[torch.arange(B, device=device), best_idx]
+
+    # Refit with the inlier set (per-batch, since the inlier counts vary).
+    H_inlier_list = []
+    for b in range(B):
+        mask = best_inlier_mask[b]
+        in_src = src_pts[b][mask]
+        in_dst = dst_pts[b][mask]
+        in_w = confident_weight[b][mask]
+        if in_src.shape[0] < 4:
+            # Fall back to identity when RANSAC fails to find enough inliers.
+            H_inlier_list.append(torch.eye(3, device=device, dtype=src_pts.dtype))
+            continue
+        sorted_w = torch.argsort(in_w, descending=True)
+        if len(sorted_w) > max_inlier_num:
+            keep = max(int(len(sorted_w) * 0.95), max_inlier_num)
+            sorted_w = sorted_w[:keep][torch.randperm(keep, device=device)[:max_inlier_num]]
+        H_inlier_list.append(
+            _find_homography_weighted_lsq(in_src[sorted_w], in_dst[sorted_w], in_w[sorted_w])
+        )
+    return torch.stack(H_inlier_list, dim=0)
+
+
+# -----------------------------------------------------------------------------
+# Camera-ray utilities
+# -----------------------------------------------------------------------------
+
+
+def _unproject_identity(num_y: int, num_x: int, B: int, S: int, device, dtype) -> torch.Tensor:
+    """Camera-space unit rays for an identity intrinsic on a 2x2 image plane."""
+    dx = 1.0 / num_x
+    dy = 1.0 / num_y
+    # Centered camera-space coords directly (skip the K^-1 step since it's
+    # just a translation by -1 on x and y when K is identity-with-center=1).
+    y = torch.linspace(-(1 - dy), (1 - dy), num_y, device=device, dtype=dtype)
+    x = torch.linspace(-(1 - dx), (1 - dx), num_x, device=device, dtype=dtype)
+    yy, xx = torch.meshgrid(y, x, indexing="ij")
+    grid = torch.stack((xx, yy), dim=-1)            # (h, w, 2)
+    grid = grid.unsqueeze(0).unsqueeze(0).expand(B, S, num_y, num_x, 2)
+    return torch.cat([grid, torch.ones_like(grid[..., :1])], dim=-1)
+
+
+def _camray_to_caminfo(
+    camray: torch.Tensor,  # (B, S, h, w, 6)
+    confidence: Optional[torch.Tensor] = None,  # (B, S, h, w)
+    reproj_threshold: float = 0.2,
+) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor]:
+    """Convert per-pixel camera rays to per-view (R, T, focal, principal)."""
+    if confidence is None:
+        confidence = torch.ones_like(camray[..., 0])
+    B, S, h, w, _ = camray.shape
+    device = camray.device
+    dtype = camray.dtype
+
+    rays_target = camray[..., :3]                           # (B, S, h, w, 3)
+    rays_origin = _unproject_identity(h, w, B, S, device, dtype)
+
+    # Flatten (B*S, h*w, *) for the RANSAC routine.
+    rays_target = rays_target.flatten(0, 1).flatten(1, 2)
+    rays_origin = rays_origin.flatten(0, 1).flatten(1, 2)
+    weights = confidence.flatten(0, 1).flatten(1, 2).clone()
+
+    # Project to 2D in homogeneous form (the upstream calls this "perspective division").
+    z_thresh = 1e-4
+    mask = (rays_target[:, :, 2].abs() > z_thresh) & (rays_origin[:, :, 2].abs() > z_thresh)
+    weights = torch.where(mask, weights, torch.zeros_like(weights))
+    src = rays_origin.clone()
+    dst = rays_target.clone()
+    src[..., 0] = torch.where(mask, src[..., 0] / src[..., 2], src[..., 0])
+    src[..., 1] = torch.where(mask, src[..., 1] / src[..., 2], src[..., 1])
+    dst[..., 0] = torch.where(mask, dst[..., 0] / dst[..., 2], dst[..., 0])
+    dst[..., 1] = torch.where(mask, dst[..., 1] / dst[..., 2], dst[..., 1])
+    src = src[..., :2]
+    dst = dst[..., :2]
+
+    N = src.shape[1]
+    n_iter = 100
+    sample_ratio = 0.3
+    num_sample_for_ransac = 8
+    n_sample = max(num_sample_for_ransac, int(N * sample_ratio))
+    rand_idx = torch.stack(
+        [torch.randperm(n_sample, device=device)[:num_sample_for_ransac] for _ in range(n_iter)],
+        dim=0,
+    )
+
+    # Chunk along the view axis to keep peak memory predictable.
+    chunk = 2
+    A_list = []
+    for i in range(0, src.shape[0], chunk):
+        A = _ransac_find_homography_weighted_batched(
+            src[i:i + chunk], dst[i:i + chunk], weights[i:i + chunk],
+            n_sample=n_sample, n_iter=n_iter,
+            num_sample_for_ransac=num_sample_for_ransac,
+            reproj_threshold=reproj_threshold,
+            rand_sample_iters_idx=rand_idx,
+            max_inlier_num=8000,
+        )
+        # Flip sign on dets that come out < 0 (so that the QL produces a
+        # right-handed rotation). ``det`` lacks fp16/bf16 CUDA kernels, so
+        # do the comparison in fp32.
+        flip = torch.linalg.det(A.float()) < 0
+        A = torch.where(flip[:, None, None], -A, A)
+        A_list.append(A)
+    A = torch.cat(A_list, dim=0)                            # (B*S, 3, 3)
+
+    R_list, f_list, pp_list = [], [], []
+    for i in range(A.shape[0]):
+        R, L = _ql_decomposition(A[i])
+        L = L / L[2][2]
+        f_list.append(torch.stack((L[0][0], L[1][1])))
+        pp_list.append(torch.stack((L[2][0], L[2][1])))
+        R_list.append(R)
+    R = torch.stack(R_list).reshape(B, S, 3, 3)
+    focal = torch.stack(f_list).reshape(B, S, 2)
+    pp = torch.stack(pp_list).reshape(B, S, 2)
+
+    # Translation: confidence-weighted average of camray direction(s).
+    cf = confidence.flatten(0, 1).flatten(1, 2)
+    T = (camray.flatten(0, 1).flatten(1, 2)[..., 3:] * cf.unsqueeze(-1)).sum(dim=1)
+    T = T / cf.sum(dim=-1, keepdim=True)
+    T = T.reshape(B, S, 3)
+
+    # Match upstream output convention: focal -> 1/focal, pp + 1.
+    return R, T, 1.0 / focal, pp + 1.0
+
+
+def get_extrinsic_from_camray(
+    camray: torch.Tensor,  # (B, S, h, w, 6)
+    conf: torch.Tensor,  # (B, S, h, w, 1) or (B, S, h, w)
+    patch_size_y: int,
+    patch_size_x: int,
+) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor]:
+    """Wrap a 4x4 extrinsic + per-view focal + principal-point output."""
+    if conf.ndim == 5 and conf.shape[-1] == 1:
+        conf = conf.squeeze(-1)
+    R, T, focal, pp = _camray_to_caminfo(camray, confidence=conf)
+    extr = torch.cat([R, T.unsqueeze(-1)], dim=-1)           # (B, S, 3, 4)
+    homo_row = torch.tensor([0, 0, 0, 1], dtype=R.dtype, device=R.device)
+    homo_row = homo_row.view(1, 1, 1, 4).expand(R.shape[0], R.shape[1], 1, 4)
+    extr = torch.cat([extr, homo_row], dim=-2)               # (B, S, 4, 4)
+    return extr, focal, pp

comfy/ldm/depth_anything_3/reference_view_selector.py

@@ -0,0 +1,87 @@
+"""Reference-view selection for the multi-view path of Depth Anything 3."""
+
+from __future__ import annotations
+
+from typing import Literal
+
+import torch
+
+
+RefViewStrategy = Literal["first", "middle", "saddle_balanced", "saddle_sim_range"]
+
+
+# Per the upstream constants module: ``THRESH_FOR_REF_SELECTION = 3``.
+# Reference selection only runs when there are at least this many views.
+THRESH_FOR_REF_SELECTION: int = 3
+
+
+def select_reference_view(x: torch.Tensor, strategy: RefViewStrategy = "saddle_balanced") -> torch.Tensor:
+    """Pick a reference view index per batch element."""
+    B, S, _, _ = x.shape
+    if S <= 1:
+        return torch.zeros(B, dtype=torch.long, device=x.device)
+    if strategy == "first":
+        return torch.zeros(B, dtype=torch.long, device=x.device)
+    if strategy == "middle":
+        return torch.full((B,), S // 2, dtype=torch.long, device=x.device)
+
+    # Feature-based strategies: normalised cls/cam token per view.
+    img_class_feat = x[:, :, 0] / x[:, :, 0].norm(dim=-1, keepdim=True)  # (B,S,C)
+
+    if strategy == "saddle_balanced":
+        sim = torch.matmul(img_class_feat, img_class_feat.transpose(1, 2))  # (B,S,S)
+        sim_no_diag = sim - torch.eye(S, device=sim.device).unsqueeze(0)
+        sim_score = sim_no_diag.sum(dim=-1) / (S - 1)               # (B,S)
+        feat_norm = x[:, :, 0].norm(dim=-1)                          # (B,S)
+        feat_var = img_class_feat.var(dim=-1)                        # (B,S)
+
+        def _normalize(metric):
+            mn = metric.min(dim=1, keepdim=True).values
+            mx = metric.max(dim=1, keepdim=True).values
+            return (metric - mn) / (mx - mn + 1e-8)
+
+        sim_n, norm_n, var_n = _normalize(sim_score), _normalize(feat_norm), _normalize(feat_var)
+        balance = (sim_n - 0.5).abs() + (norm_n - 0.5).abs() + (var_n - 0.5).abs()
+        return balance.argmin(dim=1)
+
+    if strategy == "saddle_sim_range":
+        sim = torch.matmul(img_class_feat, img_class_feat.transpose(1, 2))
+        sim_no_diag = sim - torch.eye(S, device=sim.device).unsqueeze(0)
+        sim_max = sim_no_diag.max(dim=-1).values
+        sim_min = sim_no_diag.min(dim=-1).values
+        return (sim_max - sim_min).argmax(dim=1)
+
+    raise ValueError(
+        f"Unknown reference view selection strategy: {strategy!r}. "
+        f"Must be one of: 'first', 'middle', 'saddle_balanced', 'saddle_sim_range'"
+    )
+
+
+def reorder_by_reference(x: torch.Tensor, b_idx: torch.Tensor) -> torch.Tensor:
+    """Reorder x so the reference view is at position 0 in axis S."""
+    B, S = x.shape[0], x.shape[1]
+    if S <= 1:
+        return x
+    positions = torch.arange(S, device=x.device).unsqueeze(0).expand(B, -1)
+    b_idx_exp = b_idx.unsqueeze(1)
+    reorder = torch.where(
+        (positions > 0) & (positions <= b_idx_exp),
+        positions - 1,
+        positions,
+    )
+    reorder[:, 0] = b_idx
+    batch = torch.arange(B, device=x.device).unsqueeze(1)
+    return x[batch, reorder]
+
+
+def restore_original_order(x: torch.Tensor, b_idx: torch.Tensor) -> torch.Tensor:
+    """Inverse of reorder_by_reference."""
+    B, S = x.shape[0], x.shape[1]
+    if S <= 1:
+        return x
+    target_positions = torch.arange(S, device=x.device).unsqueeze(0).expand(B, -1)
+    b_idx_exp = b_idx.unsqueeze(1)
+    restore = torch.where(target_positions < b_idx_exp, target_positions + 1, target_positions)
+    restore = torch.scatter(restore, dim=1, index=b_idx_exp, src=torch.zeros_like(b_idx_exp))
+    batch = torch.arange(B, device=x.device).unsqueeze(1)
+    return x[batch, restore]

comfy/ldm/depth_anything_3/transform.py

@@ -0,0 +1,160 @@
+"""Geometry / camera transform helpers for Depth Anything 3."""
+
+from __future__ import annotations
+
+from typing import Tuple
+
+import torch
+import torch.nn.functional as F
+
+
+# -----------------------------------------------------------------------------
+# Affine 4x4 helpers
+# -----------------------------------------------------------------------------
+
+
+def as_homogeneous(ext: torch.Tensor) -> torch.Tensor:
+    """Promote (...,3,4) extrinsics to (...,4,4) homogeneous form. No-op when the input is already ``(...,4,4)``."""
+    if ext.shape[-2:] == (4, 4):
+        return ext
+    if ext.shape[-2:] == (3, 4):
+        ones = torch.zeros_like(ext[..., :1, :4])
+        ones[..., 0, 3] = 1.0
+        return torch.cat([ext, ones], dim=-2)
+    raise ValueError(f"Invalid affine shape: {ext.shape}")
+
+
+def affine_inverse(A: torch.Tensor) -> torch.Tensor:
+    """Inverse of an affine matrix ``[R|T; 0 0 0 1]``."""
+    R = A[..., :3, :3]
+    T = A[..., :3, 3:]
+    P = A[..., 3:, :]
+    return torch.cat([torch.cat([R.mT, -R.mT @ T], dim=-1), P], dim=-2)
+
+
+# -----------------------------------------------------------------------------
+# Quaternion <-> rotation matrix (xyzw / scalar-last)
+# -----------------------------------------------------------------------------
+
+
+def _sqrt_positive_part(x: torch.Tensor) -> torch.Tensor:
+    """sqrt(max(0, x)) with a zero subgradient where x == 0."""
+    ret = torch.zeros_like(x)
+    positive_mask = x > 0
+    if torch.is_grad_enabled():
+        ret[positive_mask] = torch.sqrt(x[positive_mask])
+    else:
+        ret = torch.where(positive_mask, torch.sqrt(x), ret)
+    return ret
+
+
+def standardize_quaternion(quaternions: torch.Tensor) -> torch.Tensor:
+    """Force the real part of a unit quaternion (xyzw) to be non-negative."""
+    return torch.where(quaternions[..., 3:4] < 0, -quaternions, quaternions)
+
+
+def quat_to_mat(quaternions: torch.Tensor) -> torch.Tensor:
+    """Convert quaternions (xyzw) to (...,3,3) rotation matrices."""
+    i, j, k, r = torch.unbind(quaternions, -1)
+    two_s = 2.0 / (quaternions * quaternions).sum(-1)
+    o = torch.stack(
+        (
+            1 - two_s * (j * j + k * k),
+            two_s * (i * j - k * r),
+            two_s * (i * k + j * r),
+            two_s * (i * j + k * r),
+            1 - two_s * (i * i + k * k),
+            two_s * (j * k - i * r),
+            two_s * (i * k - j * r),
+            two_s * (j * k + i * r),
+            1 - two_s * (i * i + j * j),
+        ),
+        -1,
+    )
+    return o.reshape(quaternions.shape[:-1] + (3, 3))
+
+
+def mat_to_quat(matrix: torch.Tensor) -> torch.Tensor:
+    """Convert (...,3,3) rotation matrices to quaternions (xyzw)."""
+    if matrix.size(-1) != 3 or matrix.size(-2) != 3:
+        raise ValueError(f"Invalid rotation matrix shape {matrix.shape}.")
+
+    batch_dim = matrix.shape[:-2]
+    m00, m01, m02, m10, m11, m12, m20, m21, m22 = torch.unbind(
+        matrix.reshape(batch_dim + (9,)), dim=-1
+    )
+
+    q_abs = _sqrt_positive_part(
+        torch.stack(
+            [
+                1.0 + m00 + m11 + m22,
+                1.0 + m00 - m11 - m22,
+                1.0 - m00 + m11 - m22,
+                1.0 - m00 - m11 + m22,
+            ],
+            dim=-1,
+        )
+    )
+
+    quat_by_rijk = torch.stack(
+        [
+            torch.stack([q_abs[..., 0] ** 2, m21 - m12, m02 - m20, m10 - m01], dim=-1),
+            torch.stack([m21 - m12, q_abs[..., 1] ** 2, m10 + m01, m02 + m20], dim=-1),
+            torch.stack([m02 - m20, m10 + m01, q_abs[..., 2] ** 2, m12 + m21], dim=-1),
+            torch.stack([m10 - m01, m20 + m02, m21 + m12, q_abs[..., 3] ** 2], dim=-1),
+        ],
+        dim=-2,
+    )
+
+    flr = torch.tensor(0.1).to(dtype=q_abs.dtype, device=q_abs.device)
+    quat_candidates = quat_by_rijk / (2.0 * q_abs[..., None].max(flr))
+
+    out = quat_candidates[F.one_hot(q_abs.argmax(dim=-1), num_classes=4) > 0.5, :].reshape(
+        batch_dim + (4,)
+    )
+    # Reorder rijk -> xyzw (i.e. ijkr).
+    out = out[..., [1, 2, 3, 0]]
+    return standardize_quaternion(out)
+
+
+# -----------------------------------------------------------------------------
+# Pose-encoding <-> extrinsics + intrinsics
+# -----------------------------------------------------------------------------
+
+
+def extri_intri_to_pose_encoding(extrinsics: torch.Tensor, intrinsics: torch.Tensor, image_size_hw: Tuple[int, int]) -> torch.Tensor:
+    """Pack (extr, intr, image_size) into the 9-D pose-encoding vector.
+    extrinsics: camera-to-world (c2w) (B,S,4,4) matrices,
+    intrinsics: pixel-space (B,S,3,3) matrices,
+    image_size_hw: is a (H, W) pair.
+    """
+    R = extrinsics[..., :3, :3]
+    T = extrinsics[..., :3, 3]
+    quat = mat_to_quat(R)
+    H, W = image_size_hw
+    fov_h = 2 * torch.atan((H / 2) / intrinsics[..., 1, 1])
+    fov_w = 2 * torch.atan((W / 2) / intrinsics[..., 0, 0])
+    return torch.cat([T, quat, fov_h[..., None], fov_w[..., None]], dim=-1).float()
+
+
+def pose_encoding_to_extri_intri(pose_encoding: torch.Tensor, image_size_hw: Tuple[int, int]) -> Tuple[torch.Tensor, torch.Tensor]:
+    """Inverse of extri_intri_to_pose_encoding."""
+    T = pose_encoding[..., :3]
+    quat = pose_encoding[..., 3:7]
+    fov_h = pose_encoding[..., 7]
+    fov_w = pose_encoding[..., 8]
+    # Normalize to unit quaternion. CameraDec outputs raw values; a near-zero
+    # quaternion causes two_s = 2/norm² → inf in quat_to_mat → NaN extrinsics.
+    quat = quat / quat.norm(dim=-1, keepdim=True).clamp(min=1e-6)
+    R = quat_to_mat(quat)
+    extrinsics = torch.cat([R, T[..., None]], dim=-1)
+    H, W = image_size_hw
+    fy = (H / 2.0) / torch.clamp(torch.tan(fov_h / 2.0), 1e-6)
+    fx = (W / 2.0) / torch.clamp(torch.tan(fov_w / 2.0), 1e-6)
+    intrinsics = torch.zeros(pose_encoding.shape[:2] + (3, 3), device=pose_encoding.device, dtype=pose_encoding.dtype)
+    intrinsics[..., 0, 0] = fx
+    intrinsics[..., 1, 1] = fy
+    intrinsics[..., 0, 2] = W / 2
+    intrinsics[..., 1, 2] = H / 2
+    intrinsics[..., 2, 2] = 1.0
+    return extrinsics, intrinsics

comfy/ldm/ernie/model.py

@@ -5,6 +5,7 @@ import torch.nn.functional as F
 
 from comfy.ldm.modules.attention import optimized_attention
 import comfy.model_management
+import comfy.quant_ops
 
 def rope(pos: torch.Tensor, dim: int, theta: int) -> torch.Tensor:
     assert dim % 2 == 0
@@ -19,15 +20,6 @@ def rope(pos: torch.Tensor, dim: int, theta: int) -> torch.Tensor:
     out = torch.stack([torch.cos(out), torch.sin(out)], dim=0)
     return out.to(dtype=torch.float32, device=pos.device)
 
-def apply_rotary_emb(x_in: torch.Tensor, freqs_cis: torch.Tensor) -> torch.Tensor:
-    rot_dim = freqs_cis.shape[-1]
-    x, x_pass = x_in[..., :rot_dim], x_in[..., rot_dim:]
-    cos_ = freqs_cis[0]
-    sin_ = freqs_cis[1]
-    x1, x2 = x.chunk(2, dim=-1)
-    x_rotated = torch.cat((-x2, x1), dim=-1)
-    return torch.cat((x * cos_ + x_rotated * sin_, x_pass), dim=-1)
-
 class ErnieImageEmbedND3(nn.Module):
     def __init__(self, dim: int, theta: int, axes_dim: tuple):
         super().__init__()
@@ -37,8 +29,16 @@ class ErnieImageEmbedND3(nn.Module):
 
     def forward(self, ids: torch.Tensor) -> torch.Tensor:
         emb = torch.cat([rope(ids[..., i], self.axes_dim[i], self.theta) for i in range(3)], dim=-1)
-        emb = emb.unsqueeze(3)  # [2, B, S, 1, head_dim//2]
-        return torch.stack([emb, emb], dim=-1).reshape(*emb.shape[:-1], -1)  # [B, S, 1, head_dim]
+        cos_ = emb[0]
+        sin_ = emb[1]
+        N = cos_.shape[-1]
+        half = N // 2
+        cos_top = cos_[..., :half].repeat_interleave(2, dim=-1)
+        sin_top = sin_[..., :half].repeat_interleave(2, dim=-1)
+        cos_bot = cos_[..., half:].repeat_interleave(2, dim=-1)
+        sin_bot = sin_[..., half:].repeat_interleave(2, dim=-1)
+        rot = torch.stack([cos_top, -sin_top, sin_bot, cos_bot], dim=-1)
+        return rot.reshape(*rot.shape[:-1], 2, 2).unsqueeze(2)
 
 class ErnieImagePatchEmbedDynamic(nn.Module):
     def __init__(self, in_channels: int, embed_dim: int, patch_size: int, operations, device=None, dtype=None):
@@ -115,8 +115,7 @@ class ErnieImageAttention(nn.Module):
         key = self.norm_k(key)
 
         if image_rotary_emb is not None:
-            query = apply_rotary_emb(query, image_rotary_emb)
-            key = apply_rotary_emb(key, image_rotary_emb)
+            query, key = comfy.quant_ops.ck.apply_rope_split_half(query, key, image_rotary_emb)
 
         q_flat = query.reshape(B, S, -1)
         k_flat = key.reshape(B, S, -1)
@@ -274,7 +273,7 @@ class ErnieImageModel(nn.Module):
 
         image_ids = image_ids.view(1, N_img, 3).expand(B, -1, -1)
 
-        rotary_pos_emb = self.pos_embed(torch.cat([image_ids, text_ids], dim=1)).to(x.dtype)
+        rotary_pos_emb = self.pos_embed(torch.cat([image_ids, text_ids], dim=1))
         del image_ids, text_ids
 
         sample = self.time_proj(timesteps).to(dtype)

comfy/ldm/flux/math.py

@@ -4,7 +4,7 @@ from torch import Tensor
 
 from comfy.ldm.modules.attention import optimized_attention
 import comfy.model_management
-import logging
+import comfy.quant_ops
 
 
 def attention(q: Tensor, k: Tensor, v: Tensor, pe: Tensor, mask=None, transformer_options={}) -> Tensor:
@@ -44,21 +44,15 @@ def _apply_rope(xq: Tensor, xk: Tensor, freqs_cis: Tensor):
     return apply_rope1(xq, freqs_cis), apply_rope1(xk, freqs_cis)
 
 
-try:
-    import comfy.quant_ops
-    q_apply_rope = comfy.quant_ops.ck.apply_rope
-    q_apply_rope1 = comfy.quant_ops.ck.apply_rope1
-    def apply_rope(xq, xk, freqs_cis):
-        if comfy.model_management.in_training:
-            return _apply_rope(xq, xk, freqs_cis)
-        else:
-            return apply_rope1(xq, freqs_cis), apply_rope1(xk, freqs_cis)
-    def apply_rope1(x, freqs_cis):
-        if comfy.model_management.in_training:
-            return _apply_rope1(x, freqs_cis)
-        else:
-            return q_apply_rope1(x, freqs_cis)
-except:
-    logging.warning("No comfy kitchen, using old apply_rope functions.")
-    apply_rope = _apply_rope
-    apply_rope1 = _apply_rope1
+def apply_rope(xq, xk, freqs_cis):
+    if comfy.model_management.in_training:
+        return _apply_rope(xq, xk, freqs_cis)
+    else:
+        return comfy.quant_ops.ck.apply_rope(xq, xk, freqs_cis)
+
+
+def apply_rope1(x, freqs_cis):
+    if comfy.model_management.in_training:
+        return _apply_rope1(x, freqs_cis)
+    else:
+        return comfy.quant_ops.ck.apply_rope1(x, freqs_cis)

comfy/ldm/ideogram4/model.py

@@ -0,0 +1,297 @@
+"""
+The Ideogram 4 transformer is a NextDiT/Lumina2-family single-stream model
+consumes Qwen3-VL hidden-state features (concatenated from 13 layers -> 53248 dims)
+packs ``[text tokens, image tokens]`` into one sequence with block-diagonal segment attention and 3D interleaved MRoPE.
+"""
+
+from __future__ import annotations
+
+import math
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+import comfy.patcher_extension
+from comfy.ldm.lumina.model import FeedForward
+from comfy.ldm.modules.attention import optimized_attention_masked
+from comfy.text_encoders.llama import apply_rope, precompute_freqs_cis
+
+# Per-token role indicators
+SEQUENCE_PADDING_INDICATOR = -1
+OUTPUT_IMAGE_INDICATOR = 2
+LLM_TOKEN_INDICATOR = 3
+# Image grid coordinates are offset so they never collide with text positions
+IMAGE_POSITION_OFFSET = 65536
+
+
+class Ideogram4Attention(nn.Module):
+    def __init__(self, hidden_size, num_heads, eps=1e-5, dtype=None, device=None, operations=None):
+        super().__init__()
+        self.num_heads = num_heads
+        self.head_dim = hidden_size // num_heads
+        self.hidden_size = hidden_size
+
+        self.qkv = operations.Linear(hidden_size, hidden_size * 3, bias=False, dtype=dtype, device=device)
+        self.norm_q = operations.RMSNorm(self.head_dim, eps=eps, elementwise_affine=True, dtype=dtype, device=device)
+        self.norm_k = operations.RMSNorm(self.head_dim, eps=eps, elementwise_affine=True, dtype=dtype, device=device)
+        self.o = operations.Linear(hidden_size, hidden_size, bias=False, dtype=dtype, device=device)
+
+    def forward(self, x, attn_mask, freqs_cis, transformer_options={}):
+        batch_size, seq_len, _ = x.shape
+        qkv = self.qkv(x).view(batch_size, seq_len, 3, self.num_heads, self.head_dim)
+        q, k, v = qkv.unbind(dim=2)
+
+        q = self.norm_q(q)
+        k = self.norm_k(k)
+
+        # (B, heads, L, head_dim)
+        q = q.transpose(1, 2)
+        k = k.transpose(1, 2)
+        v = v.transpose(1, 2)
+
+        q, k = apply_rope(q, k, freqs_cis)
+
+        out = optimized_attention_masked(q, k, v, self.num_heads, attn_mask, skip_reshape=True, transformer_options=transformer_options)
+        return self.o(out)
+
+
+class Ideogram4TransformerBlock(nn.Module):
+    def __init__(self, hidden_size, intermediate_size, num_heads, norm_eps, adaln_dim, dtype=None, device=None, operations=None):
+        super().__init__()
+        self.attention = Ideogram4Attention(hidden_size, num_heads, eps=1e-5, dtype=dtype, device=device, operations=operations)
+        self.feed_forward = FeedForward(
+            dim=hidden_size, hidden_dim=intermediate_size, multiple_of=1, ffn_dim_multiplier=None,
+            operation_settings={"operations": operations, "dtype": dtype, "device": device},
+        )
+
+        self.attention_norm1 = operations.RMSNorm(hidden_size, eps=norm_eps, elementwise_affine=True, dtype=dtype, device=device)
+        self.ffn_norm1 = operations.RMSNorm(hidden_size, eps=norm_eps, elementwise_affine=True, dtype=dtype, device=device)
+        self.attention_norm2 = operations.RMSNorm(hidden_size, eps=norm_eps, elementwise_affine=True, dtype=dtype, device=device)
+        self.ffn_norm2 = operations.RMSNorm(hidden_size, eps=norm_eps, elementwise_affine=True, dtype=dtype, device=device)
+
+        self.adaln_modulation = operations.Linear(adaln_dim, 4 * hidden_size, bias=True, dtype=dtype, device=device)
+
+    def forward(self, x, attn_mask, freqs_cis, adaln_input, transformer_options={}):
+        mod = self.adaln_modulation(adaln_input)
+        scale_msa, gate_msa, scale_mlp, gate_mlp = mod.chunk(4, dim=-1)
+        gate_msa = torch.tanh(gate_msa)
+        gate_mlp = torch.tanh(gate_mlp)
+        scale_msa = 1.0 + scale_msa
+        scale_mlp = 1.0 + scale_mlp
+
+        attn_out = self.attention(self.attention_norm1(x) * scale_msa, attn_mask, freqs_cis, transformer_options=transformer_options)
+        x = x + gate_msa * self.attention_norm2(attn_out)
+        x = x + gate_mlp * self.ffn_norm2(self.feed_forward(self.ffn_norm1(x) * scale_mlp))
+        return x
+
+
+def _sinusoidal_embedding(t, dim, scale=1e4):
+    t = t.to(torch.float32)
+    half = dim // 2
+    freq = math.log(scale) / (half - 1)
+    freq = torch.exp(torch.arange(half, dtype=torch.float32, device=t.device) * -freq)
+    emb = t.unsqueeze(-1) * freq
+    emb = torch.cat([torch.sin(emb), torch.cos(emb)], dim=-1)
+    if dim % 2 == 1:
+        emb = F.pad(emb, (0, 1))
+    return emb
+
+
+class Ideogram4EmbedScalar(nn.Module):
+    def __init__(self, dim, input_range=(0.0, 1.0), dtype=None, device=None, operations=None):
+        super().__init__()
+        self.dim = dim
+        self.range_min, self.range_max = input_range
+        self.mlp_in = operations.Linear(dim, dim, bias=True, dtype=dtype, device=device)
+        self.mlp_out = operations.Linear(dim, dim, bias=True, dtype=dtype, device=device)
+
+    def forward(self, x, dtype):
+        x = x.to(torch.float32)
+        scaled = 1e4 * (x - self.range_min) / (self.range_max - self.range_min)
+        emb = _sinusoidal_embedding(scaled, self.dim)
+        emb = emb.to(dtype)
+        emb = F.silu(self.mlp_in(emb))
+        return self.mlp_out(emb)
+
+
+class Ideogram4FinalLayer(nn.Module):
+    def __init__(self, hidden_size, out_channels, adaln_dim, dtype=None, device=None, operations=None):
+        super().__init__()
+        self.norm_final = operations.LayerNorm(hidden_size, eps=1e-6, elementwise_affine=False, dtype=dtype, device=device)
+        self.linear = operations.Linear(hidden_size, out_channels, bias=True, dtype=dtype, device=device)
+        self.adaln_modulation = operations.Linear(adaln_dim, hidden_size, bias=True, dtype=dtype, device=device)
+
+    def forward(self, x, c):
+        scale = 1.0 + self.adaln_modulation(F.silu(c))
+        return self.linear(self.norm_final(x) * scale)
+
+
+class Ideogram4Transformer(nn.Module):
+    """A single Ideogram 4 backbone operating on a packed token sequence."""
+
+    def __init__(self, emb_dim, num_layers, num_heads, intermediate_size, adaln_dim,
+                 in_channels, llm_features_dim, rope_theta, mrope_section, norm_eps,
+                 dtype=None, device=None, operations=None):
+        super().__init__()
+        self.head_dim = emb_dim // num_heads
+        self.rope_theta = rope_theta
+        self.mrope_section = tuple(mrope_section)
+
+        self.input_proj = operations.Linear(in_channels, emb_dim, bias=True, dtype=dtype, device=device)
+        self.llm_cond_norm = operations.RMSNorm(llm_features_dim, eps=1e-6, elementwise_affine=True, dtype=dtype, device=device)
+        self.llm_cond_proj = operations.Linear(llm_features_dim, emb_dim, bias=True, dtype=dtype, device=device)
+        self.t_embedding = Ideogram4EmbedScalar(emb_dim, input_range=(0.0, 1.0), dtype=dtype, device=device, operations=operations)
+        self.adaln_proj = operations.Linear(emb_dim, adaln_dim, bias=True, dtype=dtype, device=device)
+
+        self.embed_image_indicator = operations.Embedding(2, emb_dim, dtype=dtype, device=device)
+
+        self.layers = nn.ModuleList([
+            Ideogram4TransformerBlock(emb_dim, intermediate_size, num_heads, norm_eps, adaln_dim,
+                                      dtype=dtype, device=device, operations=operations)
+            for _ in range(num_layers)
+        ])
+
+        self.final_layer = Ideogram4FinalLayer(emb_dim, in_channels, adaln_dim, dtype=dtype, device=device, operations=operations)
+
+    def _backbone(self, llm_features, x, t, position_ids, attn_mask, indicator, transformer_options={}):
+        indicator = indicator.to(torch.long)
+        output_image_mask = (indicator == OUTPUT_IMAGE_INDICATOR).to(x.dtype).unsqueeze(-1)
+
+        x = x * output_image_mask
+        h = self.input_proj(x) * output_image_mask
+
+        t_cond = self.t_embedding(t, dtype=x.dtype)
+        if t.dim() == 1:
+            t_cond = t_cond.unsqueeze(1)
+        adaln_input = F.silu(self.adaln_proj(t_cond))
+
+        # h is zero on the text rows (content lives only on image rows), add writes the text features in place
+        if llm_features is not None:
+            L_text = llm_features.shape[1]
+            text_mask = (indicator[:, :L_text] == LLM_TOKEN_INDICATOR).to(x.dtype).unsqueeze(-1)
+            llm = self.llm_cond_norm(llm_features * text_mask)
+            llm = self.llm_cond_proj(llm) * text_mask
+            h[:, :L_text] = h[:, :L_text] + llm
+
+        h = h + self.embed_image_indicator((indicator == OUTPUT_IMAGE_INDICATOR).to(torch.long), out_dtype=h.dtype)
+
+        # Qwen3-VL interleaved MRoPE; position_ids (B, L, 3) -> (3, L) (same across batch).
+        freqs_cis = precompute_freqs_cis(
+            self.head_dim, position_ids[0].transpose(0, 1), self.rope_theta,
+            rope_dims=self.mrope_section, interleaved_mrope=True, device=position_ids.device,
+        )
+
+        if attn_mask is not None and attn_mask.dtype == torch.bool:
+            attn_mask = torch.zeros_like(attn_mask, dtype=h.dtype).masked_fill_(~attn_mask, -torch.finfo(h.dtype).max)
+
+        for layer in self.layers:
+            h = layer(h, attn_mask, freqs_cis, adaln_input, transformer_options=transformer_options)
+
+        return self.final_layer(h, adaln_input)
+
+
+class Ideogram4Transformer2DModel(Ideogram4Transformer):
+    """Ideogram 4 single-stream DiT.
+
+    Runs a packed ``[text, image]`` sequence when text context is supplied, or an image-only sequence when ``context is None``.
+    """
+
+    def __init__(self, image_model=None, in_channels=128, num_layers=34, num_attention_heads=18, attention_head_dim=256, intermediate_size=12288,
+                 adaln_dim=512, llm_features_dim=53248, rope_theta=5000000, mrope_section=(24, 20, 20), norm_eps=1e-5,
+                 dtype=None, device=None, operations=None, **kwargs):
+        emb_dim = num_attention_heads * attention_head_dim
+        super().__init__(
+            emb_dim=emb_dim, num_layers=num_layers, num_heads=num_attention_heads,
+            intermediate_size=intermediate_size, adaln_dim=adaln_dim, in_channels=in_channels,
+            llm_features_dim=llm_features_dim, rope_theta=rope_theta, mrope_section=mrope_section,
+            norm_eps=norm_eps, dtype=dtype, device=device, operations=operations)
+        self.dtype = dtype
+        self.in_channels = in_channels
+        self.out_channels = in_channels
+        # 128-dim token = patch (2x2) * ae_channels (32).
+        self.patch_size = 2
+        self.ae_channels = in_channels // (self.patch_size * self.patch_size)
+
+    def _img_to_tokens(self, x):
+        B, C, gh, gw = x.shape
+        x = x.view(B, self.ae_channels, self.patch_size, self.patch_size, gh, gw)
+        x = x.permute(0, 4, 5, 2, 3, 1)  # (B, gh, gw, pi, pj, c)
+        return x.reshape(B, gh * gw, C)
+
+    def _tokens_to_img(self, tokens, gh, gw):
+        B = tokens.shape[0]
+        C = tokens.shape[-1]
+        x = tokens.reshape(B, gh, gw, self.patch_size, self.patch_size, self.ae_channels)
+        x = x.permute(0, 5, 3, 4, 1, 2)  # (B, c, pi, pj, gh, gw)
+        return x.reshape(B, C, gh, gw)
+
+    def _image_position_ids(self, gh, gw, device):
+        h_idx = torch.arange(gh, device=device).view(-1, 1).expand(gh, gw).reshape(-1)
+        w_idx = torch.arange(gw, device=device).view(1, -1).expand(gh, gw).reshape(-1)
+        t_idx = torch.zeros_like(h_idx)
+        return torch.stack([t_idx, h_idx, w_idx], dim=1) + IMAGE_POSITION_OFFSET  # (L_img, 3)
+
+    def _run_conditional(self, x_chunk, context_chunk, attn_mask_chunk, t_chunk, gh, gw, transformer_options):
+        B = x_chunk.shape[0]
+        device = x_chunk.device
+        img_tokens = self._img_to_tokens(x_chunk)
+        L_img = img_tokens.shape[1]
+        L_text = context_chunk.shape[1]
+        L = L_text + L_img
+        latent_dim = img_tokens.shape[-1]
+
+        x_full = torch.zeros(B, L, latent_dim, dtype=img_tokens.dtype, device=device)
+        x_full[:, L_text:] = img_tokens
+
+        text_pos = torch.arange(L_text, device=device).view(-1, 1).expand(L_text, 3)
+        img_pos = self._image_position_ids(gh, gw, device)
+        position_ids = torch.cat([text_pos, img_pos], dim=0).unsqueeze(0).expand(B, L, 3)
+
+        indicator = torch.empty(B, L, dtype=torch.long, device=device)
+        indicator[:, :L_text] = LLM_TOKEN_INDICATOR
+        indicator[:, L_text:] = OUTPUT_IMAGE_INDICATOR
+
+        attn_mask = None
+        if attn_mask_chunk is not None:
+            segment_ids = torch.ones(B, L, dtype=torch.long, device=device)
+            pad = (attn_mask_chunk == 0)
+            segment_ids[:, :L_text][pad] = SEQUENCE_PADDING_INDICATOR
+            indicator[:, :L_text][pad] = 0
+            # Block-diagonal mask from segment ids: (B, 1, L, L), True = attend.
+            attn_mask = (segment_ids.unsqueeze(2) == segment_ids.unsqueeze(1)).unsqueeze(1)
+
+        out = self._backbone(context_chunk, x_full, t_chunk, position_ids, attn_mask, indicator,
+                             transformer_options=transformer_options)
+        return self._tokens_to_img(out[:, L_text:], gh, gw)
+
+    def _run_image_only(self, x_chunk, t_chunk, gh, gw, transformer_options):
+        B = x_chunk.shape[0]
+        device = x_chunk.device
+        img_tokens = self._img_to_tokens(x_chunk)
+        L_img = img_tokens.shape[1]
+
+        position_ids = self._image_position_ids(gh, gw, device).unsqueeze(0).expand(B, L_img, 3)
+        indicator = torch.full((B, L_img), OUTPUT_IMAGE_INDICATOR, dtype=torch.long, device=device)
+
+        # Image-only sequence is a single segment -> no mask, full attention, no LLM context.
+        out = self._backbone(None, img_tokens, t_chunk, position_ids, None, indicator, transformer_options=transformer_options)
+        return self._tokens_to_img(out, gh, gw)
+
+    def forward(self, x, timesteps, context=None, attention_mask=None, transformer_options={}, **kwargs):
+        return comfy.patcher_extension.WrapperExecutor.new_class_executor(
+            self._forward,
+            self,
+            comfy.patcher_extension.get_all_wrappers(comfy.patcher_extension.WrappersMP.DIFFUSION_MODEL, transformer_options),
+        ).execute(x, timesteps, context, attention_mask, transformer_options, **kwargs)
+
+    def _forward(self, x, timesteps, context=None, attention_mask=None, transformer_options={}, **kwargs):
+        bs, c, gh, gw = x.shape
+
+        timesteps = 1.0 - timesteps
+
+        # unconditional pass
+        if context is None:
+            return -self._run_image_only(x, timesteps, gh, gw, transformer_options)
+
+        return -self._run_conditional(x, context, attention_mask, timesteps, gh, gw, transformer_options)

comfy/ldm/lightricks/model.py

@@ -1085,7 +1085,7 @@ class LTXVModel(LTXBaseModel):
         )
 
         grid_mask = None
-        if keyframe_idxs is not None:
+        if keyframe_idxs is not None and keyframe_idxs.shape[2] > 0:
             additional_args.update({ "orig_patchified_shape": list(x.shape)})
             denoise_mask = self.patchifier.patchify(denoise_mask)[0]
             grid_mask = ~torch.any(denoise_mask < 0, dim=-1)[0]
@@ -1330,7 +1330,7 @@ class LTXVModel(LTXBaseModel):
         x = x * (1 + scale) + shift
         x = self.proj_out(x)
 
-        if keyframe_idxs is not None:
+        if keyframe_idxs is not None and keyframe_idxs.shape[2] > 0:
             grid_mask = kwargs["grid_mask"]
             orig_patchified_shape = kwargs["orig_patchified_shape"]
             full_x = torch.zeros(orig_patchified_shape, dtype=x.dtype, device=x.device)

comfy/ldm/omnigen/omnigen2.py

@@ -8,6 +8,7 @@ import torch.nn.functional as F
 from einops import rearrange, repeat
 from comfy.ldm.lightricks.model import Timesteps
 from comfy.ldm.flux.layers import EmbedND
+from comfy.ldm.flux.math import apply_rope1
 from comfy.ldm.modules.attention import optimized_attention_masked
 import comfy.model_management
 import comfy.ldm.common_dit
@@ -17,13 +18,11 @@ def apply_rotary_emb(x, freqs_cis):
     if x.shape[1] == 0:
         return x
 
-    t_ = x.reshape(*x.shape[:-1], -1, 1, 2)
-    t_out = freqs_cis[..., 0] * t_[..., 0] + freqs_cis[..., 1] * t_[..., 1]
-    return t_out.reshape(*x.shape).to(dtype=x.dtype)
+    return apply_rope1(x, freqs_cis)
 
 
 def swiglu(x: torch.Tensor, y: torch.Tensor) -> torch.Tensor:
-    return F.silu(x) * y
+    return F.silu(x, inplace=True).mul_(y)
 
 
 class TimestepEmbedding(nn.Module):

comfy/ldm/qwen_image/model.py

@@ -51,6 +51,18 @@ class FeedForward(nn.Module):
         return hidden_states
 
 
+# Addin this back because Nunchaku custom nodes rely on it, see comment here:
+# https://github.com/Comfy-Org/ComfyUI/pull/14178#issuecomment-4640475161
+# TODO: Eventually remove this once we natively support SVDQuants
+def apply_rotary_emb(x, freqs_cis):
+    if x.shape[1] == 0:
+        return x
+
+    t_ = x.reshape(*x.shape[:-1], -1, 1, 2)
+    t_out = freqs_cis[..., 0] * t_[..., 0] + freqs_cis[..., 1] * t_[..., 1]
+    return t_out.reshape(*x.shape)
+
+
 class QwenTimestepProjEmbeddings(nn.Module):
     def __init__(self, embedding_dim, pooled_projection_dim, use_additional_t_cond=False, dtype=None, device=None, operations=None):
         super().__init__()

comfy/ldm/triposplat/gaussian.py

@@ -0,0 +1,199 @@
+# TripoSplat 3D gaussian container. Operates on already-decoded
+# tensors and exposes them as render-ready tensors (render_tensors) for the generic SPLAT type.
+import torch
+import torch.nn.functional as F
+
+import comfy.model_management
+
+
+class GaussianModel:
+    def __init__(self, aabb: list, sh_degree: int = 0, mininum_kernel_size: float = 0.0,
+                 scaling_bias: float = 0.01, opacity_bias: float = 0.1,
+                 scaling_activation: str = "exp", device=None):
+        self.sh_degree = sh_degree
+        self.mininum_kernel_size = mininum_kernel_size
+        self.scaling_bias = scaling_bias
+        self.opacity_bias = opacity_bias
+        self.device = device
+        self.aabb = torch.tensor(aabb, dtype=torch.float32, device=device)
+
+        if scaling_activation == "exp":
+            self._scaling_activation = torch.exp
+            self._inverse_scaling_activation = torch.log
+        elif scaling_activation == "softplus":
+            self._scaling_activation = F.softplus
+            self._inverse_scaling_activation = lambda x: x + torch.log(-torch.expm1(-x))
+
+        self._opacity_activation = torch.sigmoid
+        self._inverse_opacity_activation = lambda x: torch.log(x / (1 - x))
+
+        self.scale_bias = self._inverse_scaling_activation(torch.tensor(self.scaling_bias)).to(self.device)
+        self.rots_bias = torch.zeros(4, device=self.device)
+        self.rots_bias[0] = 1
+        self.opacity_bias_val = self._inverse_opacity_activation(torch.tensor(self.opacity_bias)).to(self.device)
+
+        self._storage = {}
+
+    def _get_store(self, name):
+        return self._storage.get(name)
+
+    def _set_store(self, name, value):
+        self._storage[name] = value
+
+    @property
+    def _xyz(self):
+        return self._get_store("_xyz")
+    @_xyz.setter
+    def _xyz(self, value):
+        if value is None:
+            self._set_store("_xyz", None)
+            self._set_store("xyz", None)
+            return
+        self._set_store("_xyz", value)
+        self._set_store("xyz", value * self.aabb[None, 3:] + self.aabb[None, :3])
+
+    @property
+    def get_xyz(self):
+        return self._get_store("xyz")
+
+    @property
+    def _features_dc(self):
+        return self._get_store("_features_dc")
+    @_features_dc.setter
+    def _features_dc(self, value):
+        self._set_store("_features_dc", value)
+
+    @property
+    def _opacity(self):
+        return self._get_store("_opacity")
+    @_opacity.setter
+    def _opacity(self, value):
+        if value is None:
+            self._set_store("_opacity", None)
+            self._set_store("opacity", None)
+            return
+        self._set_store("_opacity", value)
+        self._set_store("opacity", self._opacity_activation(value + self.opacity_bias_val))
+
+    @property
+    def get_opacity(self):
+        return self._get_store("opacity")
+
+    @property
+    def _scaling(self):
+        return self._get_store("_scaling")
+    @_scaling.setter
+    def _scaling(self, value):
+        if value is None:
+            self._set_store("_scaling", None)
+            self._set_store("scaling", None)
+            return
+        self._set_store("_scaling", value)
+        s = self._scaling_activation(value + self.scale_bias)
+        s = torch.square(s) + self.mininum_kernel_size ** 2
+        self._set_store("scaling", torch.sqrt(s))
+
+    @property
+    def get_scaling(self):
+        return self._get_store("scaling")
+
+    @property
+    def _rotation(self):
+        return self._get_store("_rotation")
+    @_rotation.setter
+    def _rotation(self, value):
+        self._set_store("_rotation", value)
+
+    _DEFAULT_TRANSFORM = [[1, 0, 0], [0, 0, -1], [0, 1, 0]]
+
+    def render_tensors(self):
+        # Render-ready (activated, world-space) tensors for the generic SPLAT type. The axis transform
+        # (a 3x3 rotation, object frame -> viewer Y-up) is baked into positions and rotations.
+        # Returns float tensors on the intermediate device: positions (N,3), scales (N,3) linear,
+        # rotations (N,4) wxyz, opacities (N,1) in [0,1], sh (N,K,3) coefficients.
+        xyz = self.get_xyz.float()
+        scaling = self.get_scaling.float()
+        opacity = self.get_opacity.float()
+        rotation = (self._rotation + self.rots_bias[None, :]).float()
+        sh = self._features_dc.float()  # (N, K, 3)
+        T = torch.as_tensor(self._DEFAULT_TRANSFORM, dtype=torch.float32, device=xyz.device)
+        xyz = xyz @ T.T
+        rotation = _matrix_to_quat(torch.matmul(T, _quat_to_matrix(rotation)))
+        rotation = rotation / torch.linalg.norm(rotation, dim=-1, keepdim=True)
+        out_device = comfy.model_management.intermediate_device()
+        return (
+            xyz.to(out_device).contiguous(), scaling.to(out_device).contiguous(),
+            rotation.to(out_device).contiguous(), opacity.to(out_device).contiguous(),
+            sh.to(out_device).contiguous(),
+        )
+
+
+def _quat_to_matrix(q):
+    q = q / torch.linalg.norm(q, dim=-1, keepdim=True)
+    w, x, y, z = q[:, 0], q[:, 1], q[:, 2], q[:, 3]
+    R = torch.stack([
+        1 - 2*(y*y + z*z), 2*(x*y - w*z),     2*(x*z + w*y),
+        2*(x*y + w*z),     1 - 2*(x*x + z*z), 2*(y*z - w*x),
+        2*(x*z - w*y),     2*(y*z + w*x),     1 - 2*(x*x + y*y),
+    ], dim=-1).reshape(-1, 3, 3)
+    return R
+
+
+def _matrix_to_quat(R):
+    trace = R[:, 0, 0] + R[:, 1, 1] + R[:, 2, 2]
+    q = torch.zeros((R.shape[0], 4), dtype=R.dtype, device=R.device)
+    s = torch.sqrt(torch.clamp(trace + 1, min=0)) * 2
+    q[:, 0] = 0.25 * s
+    denom = torch.where(s != 0, s, torch.ones_like(s))
+    q[:, 1] = (R[:, 2, 1] - R[:, 1, 2]) / denom
+    q[:, 2] = (R[:, 0, 2] - R[:, 2, 0]) / denom
+    q[:, 3] = (R[:, 1, 0] - R[:, 0, 1]) / denom
+    m01 = (R[:, 0, 0] >= R[:, 1, 1]) & (R[:, 0, 0] >= R[:, 2, 2]) & (s == 0)
+    s1 = torch.sqrt(torch.clamp(1 + R[:, 0, 0] - R[:, 1, 1] - R[:, 2, 2], min=0)) * 2
+    q[m01, 0] = (R[m01, 2, 1] - R[m01, 1, 2]) / s1[m01]
+    q[m01, 1] = 0.25 * s1[m01]
+    q[m01, 2] = (R[m01, 0, 1] + R[m01, 1, 0]) / s1[m01]
+    q[m01, 3] = (R[m01, 0, 2] + R[m01, 2, 0]) / s1[m01]
+    m11 = (R[:, 1, 1] > R[:, 0, 0]) & (R[:, 1, 1] >= R[:, 2, 2]) & (s == 0)
+    s2 = torch.sqrt(torch.clamp(1 + R[:, 1, 1] - R[:, 0, 0] - R[:, 2, 2], min=0)) * 2
+    q[m11, 0] = (R[m11, 0, 2] - R[m11, 2, 0]) / s2[m11]
+    q[m11, 1] = (R[m11, 0, 1] + R[m11, 1, 0]) / s2[m11]
+    q[m11, 2] = 0.25 * s2[m11]
+    q[m11, 3] = (R[m11, 1, 2] + R[m11, 2, 1]) / s2[m11]
+    m21 = (R[:, 2, 2] > R[:, 0, 0]) & (R[:, 2, 2] > R[:, 1, 1]) & (s == 0)
+    s3 = torch.sqrt(torch.clamp(1 + R[:, 2, 2] - R[:, 0, 0] - R[:, 1, 1], min=0)) * 2
+    q[m21, 0] = (R[m21, 1, 0] - R[m21, 0, 1]) / s3[m21]
+    q[m21, 1] = (R[m21, 0, 2] + R[m21, 2, 0]) / s3[m21]
+    q[m21, 2] = (R[m21, 1, 2] + R[m21, 2, 1]) / s3[m21]
+    q[m21, 3] = 0.25 * s3[m21]
+    return q / torch.linalg.norm(q, dim=-1, keepdim=True)
+
+
+def build_gaussian_models(decoder, points_pred: dict, pred: dict):
+    # Assemble GaussianModels from the elastic decoder layout. decoder is the ElasticGaussianFixedlenDecoder
+    # (carries layout / rep_config / _get_offset)
+    x = points_pred
+    offset = decoder._get_offset(pred['features'])
+    h = pred["features"]
+    ret = []
+    for i in range(h.shape[0]):
+        g = GaussianModel(
+            sh_degree=0,
+            aabb=[-0.5, -0.5, -0.5, 1.0, 1.0, 1.0],
+            mininum_kernel_size=decoder.rep_config['filter_kernel_size_3d'],
+            scaling_bias=decoder.rep_config['scaling_bias'],
+            opacity_bias=decoder.rep_config['opacity_bias'],
+            scaling_activation=decoder.rep_config['scaling_activation'],
+            device=h.device,
+        )
+        _x = x["points"][i, :, None, :]
+        for k, v in decoder.layout.items():
+            if k == '_xyz':
+                setattr(g, k, (offset[i] + _x).flatten(0, 1))
+            elif k in ('_xyz_center', '_offset_scale'):
+                continue
+            else:
+                feats = h[i][:, v['range'][0]:v['range'][1]].reshape(-1, *v['shape']).flatten(0, 1)
+                setattr(g, k, feats * decoder.rep_config['lr'][k])
+        ret.append(g)
+    return ret

comfy/ldm/triposplat/model.py

@@ -0,0 +1,326 @@
+# TripoSplat flow-matching denoiser (LatentSeqMMFlowModel). Registered as a ModelType.FLOW arch and
+# driven by the standard KSampler; jointly denoises the (B, 8192, 16) latent and a (B, 1, 5) camera token
+# carried as a 2-element nested latent.
+import numpy as np
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+import comfy.model_management
+import comfy.patcher_extension
+import comfy.rmsnorm
+from comfy.ldm.modules.attention import optimized_attention
+from comfy.ldm.flux.math import apply_rope
+
+
+class MultiHeadRMSNorm(nn.Module):
+    def __init__(self, dim, heads, dtype=None, device=None):
+        super().__init__()
+        self.gamma = nn.Parameter(torch.empty(heads, dim, dtype=dtype, device=device))
+
+    def forward(self, x):
+        x = comfy.rmsnorm.rms_norm(x)
+        return x * comfy.model_management.cast_to(self.gamma, x.dtype, x.device)
+
+
+# Positional embeddings
+
+class RePo3DRotaryEmbedding(nn.Module):
+    def __init__(self, model_channels, num_heads, head_dim, repo_hidden_ratio=0.125, max_freq=16.0,
+                 dtype=None, device=None, operations=None):
+        super().__init__()
+        self.num_heads = num_heads
+        self.head_dim = head_dim
+        repo_hidden_size = int(model_channels * repo_hidden_ratio)
+        self.norm = operations.LayerNorm(model_channels, dtype=dtype, device=device)
+        self.gate_map = operations.Linear(model_channels, repo_hidden_size, bias=False, dtype=dtype, device=device)
+        self.content_map = operations.Linear(model_channels, repo_hidden_size, bias=False, dtype=dtype, device=device)
+        self.act = nn.SiLU()
+        self.final_map = operations.Linear(repo_hidden_size, 3 * num_heads, bias=False, dtype=dtype, device=device)
+        self.dim_0 = 2 * (head_dim // 6)
+        self.dim_1 = 2 * (head_dim // 6)
+        self.dim_2 = head_dim - self.dim_0 - self.dim_1
+        dims = [self.dim_0, self.dim_1, self.dim_2]
+        freqs_list = []
+        for d in dims:
+            freq_dim = d // 2
+            freqs_list.append(torch.linspace(1.0, float(max_freq), steps=freq_dim, dtype=torch.float32))
+        self.freqs_0 = nn.Parameter(freqs_list[0])
+        self.freqs_1 = nn.Parameter(freqs_list[1])
+        self.freqs_2 = nn.Parameter(freqs_list[2])
+
+    def forward(self, hidden_states):
+        h = self.norm(hidden_states)
+        feat = self.act(self.gate_map(h)) * self.content_map(h)
+        out = self.final_map(feat)
+        B, L, _ = out.shape
+        delta_pos = out.reshape(B, L, self.num_heads, 3)
+        f0 = comfy.model_management.cast_to(self.freqs_0, torch.float32, out.device)
+        f1 = comfy.model_management.cast_to(self.freqs_1, torch.float32, out.device)
+        f2 = comfy.model_management.cast_to(self.freqs_2, torch.float32, out.device)
+        ang_0 = delta_pos[..., 0].unsqueeze(-1) * f0 * torch.pi
+        ang_1 = delta_pos[..., 1].unsqueeze(-1) * f1 * torch.pi
+        ang_2 = delta_pos[..., 2].unsqueeze(-1) * f2 * torch.pi
+        ang = torch.cat([ang_0, ang_1, ang_2], dim=-1).float()  # (B, L, heads, head_dim/2)
+        cos, sin = ang.cos(), ang.sin()
+        return torch.stack([cos, -sin, sin, cos], dim=-1).reshape(*ang.shape, 2, 2)
+
+
+class PcdAbsolutePositionEmbedder(nn.Module):
+    # Sinusoidal absolute position embedding. Two fixed schedules are used in TripoSplat:
+    # "pow2" (flow-model latent anchors) and "log2" (octree / gaussian decoders).
+    def __init__(self, channels: int, in_channels: int = 3, max_res: int = 16, schedule: str = "pow2"):
+        super().__init__()
+        self.channels = channels
+        self.in_channels = in_channels
+        self.max_res = max_res
+        self.schedule = schedule
+        self.freq_dim = channels // in_channels // 2
+
+    def _freqs(self, device):
+        if self.schedule == "pow2":
+            freqs_2exp = torch.arange(self.max_res, dtype=torch.float32, device=device)
+            res_dim = max(0, self.freq_dim - self.max_res)
+            freqs_res = (torch.arange(res_dim, dtype=torch.float32, device=device) / max(res_dim, 1) * self.max_res
+                         if res_dim > 0 else torch.empty(0, device=device))
+            freqs = torch.cat([freqs_2exp, freqs_res], dim=0)[:self.freq_dim]
+            return torch.pow(2.0, freqs) * 2.0  # *2 folds this schedule's 2*pi into the shared *pi below
+        logs = torch.linspace(0.0, float(self.max_res), steps=self.freq_dim, dtype=torch.float32, device=device)
+        return torch.pow(2.0, logs)
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        orig_dtype = x.dtype
+        x = x.float()
+        *dims, D = x.shape
+        out = torch.outer(x.reshape(-1), self._freqs(x.device)) * torch.pi
+        out = torch.cat([out.sin(), out.cos()], dim=-1).reshape(*dims, -1)
+        if out.shape[-1] < self.channels:
+            out = torch.cat([out, torch.zeros(*dims, self.channels - out.shape[-1],
+                                              device=out.device, dtype=out.dtype)], dim=-1)
+        return out.to(orig_dtype)
+
+
+def attention(q, k, v, transformer_options=None):
+    # q, k, v: (B, L, heads, dim) -> (B, L, heads, dim). Shared optimized_attention call convention.
+    out = optimized_attention(q.transpose(1, 2), k.transpose(1, 2), v.transpose(1, 2), heads=q.shape[2],
+                              skip_reshape=True, skip_output_reshape=True, low_precision_attention=False,
+                              transformer_options=transformer_options)
+    return out.transpose(1, 2)
+
+
+# Transformer building blocks
+
+class MLP(nn.Module):
+    def __init__(self, in_channels, hidden_channels, out_channels, dtype=None, device=None, operations=None):
+        super().__init__()
+        self.mlp = nn.Sequential(
+            operations.Linear(in_channels, hidden_channels, dtype=dtype, device=device),
+            nn.GELU(approximate="tanh"),
+            operations.Linear(hidden_channels, out_channels, dtype=dtype, device=device),
+        )
+
+    def forward(self, x):
+        return self.mlp(x)
+
+
+class RopeMultiHeadAttention(nn.Module):
+    def __init__(self, channels, num_heads, qkv_bias=True, qk_rms_norm=False, use_rope=False,
+                 dtype=None, device=None, operations=None):
+        super().__init__()
+        self.channels = channels
+        self.num_heads = num_heads
+        self.head_dim = channels // num_heads
+        self.qk_rms_norm = qk_rms_norm
+        self.use_rope = use_rope
+        self.qkv = operations.Linear(channels, channels * 3, bias=qkv_bias, dtype=dtype, device=device)
+        if self.qk_rms_norm:
+            self.q_norm = MultiHeadRMSNorm(self.head_dim, num_heads, dtype=dtype, device=device)
+            self.k_norm = MultiHeadRMSNorm(self.head_dim, num_heads, dtype=dtype, device=device)
+        self.out = operations.Linear(channels, channels, dtype=dtype, device=device)
+
+    def forward(self, x, rope_emb=None, transformer_options=None):
+        B, L, C = x.shape
+        qkv = self.qkv(x).reshape(B, L, 3, self.num_heads, self.head_dim)
+        q, k, v = qkv.unbind(2)
+        if self.use_rope:
+            q, k = apply_rope(q, k, rope_emb)
+        if self.qk_rms_norm:
+            q = self.q_norm(q)
+            k = self.k_norm(k)
+        h = attention(q, k, v, transformer_options)  # (B, L, heads, dim)
+        return self.out(h.reshape(B, L, C))
+
+
+class UnifiedTransformerBlock(nn.Module):
+    def __init__(self, channels, num_heads, mlp_ratio=4.0,
+                 use_rope=False, qk_rms_norm=False, qkv_bias=True,
+                 modulation=True, share_mod=False,
+                 dtype=None, device=None, operations=None):
+        super().__init__()
+        self.modulation = modulation
+        self.share_mod = share_mod
+        self.norm1 = operations.LayerNorm(channels, elementwise_affine=not modulation, eps=1e-6, dtype=dtype, device=device)
+        self.norm2 = operations.LayerNorm(channels, elementwise_affine=not modulation, eps=1e-6, dtype=dtype, device=device)
+        self.attn = RopeMultiHeadAttention(channels, num_heads=num_heads,
+                                           qkv_bias=qkv_bias, use_rope=use_rope, qk_rms_norm=qk_rms_norm,
+                                           dtype=dtype, device=device, operations=operations)
+        self.mlp = MLP(channels, int(channels * mlp_ratio), channels, dtype=dtype, device=device, operations=operations)
+        if modulation:
+            if not share_mod:
+                self.adaLN_modulation = nn.Sequential(
+                    nn.SiLU(), operations.Linear(channels, 6 * channels, bias=True, dtype=dtype, device=device))
+            self.shift_table = nn.Parameter(torch.empty(1, 6 * channels, dtype=dtype, device=device))
+
+    def forward(self, x, mod=None, rotary_emb=None, transformer_options=None):
+        if self.modulation:
+            if not self.share_mod:
+                mod = self.adaLN_modulation(mod)
+            mod = mod + comfy.model_management.cast_to(self.shift_table, mod.dtype, mod.device)
+            shift_msa, scale_msa, gate_msa, shift_mlp, scale_mlp, gate_mlp = mod.chunk(6, dim=1)
+            h = torch.addcmul(shift_msa.unsqueeze(1), self.norm1(x), 1 + scale_msa.unsqueeze(1))
+            x = torch.addcmul(x, self.attn(h, rope_emb=rotary_emb, transformer_options=transformer_options), gate_msa.unsqueeze(1))
+            h = torch.addcmul(shift_mlp.unsqueeze(1), self.norm2(x), 1 + scale_mlp.unsqueeze(1))
+            x = torch.addcmul(x, self.mlp(h), gate_mlp.unsqueeze(1))
+        else:
+            x = x + self.attn(self.norm1(x), rope_emb=rotary_emb, transformer_options=transformer_options)
+            x = x + self.mlp(self.norm2(x))
+        return x
+
+
+class TimestepEmbedder(nn.Module):
+    def __init__(self, hidden_size, frequency_embedding_size=256, dtype=None, device=None, operations=None):
+        super().__init__()
+        self.mlp = nn.Sequential(
+            operations.Linear(frequency_embedding_size, hidden_size, bias=True, dtype=dtype, device=device),
+            nn.SiLU(),
+            operations.Linear(hidden_size, hidden_size, bias=True, dtype=dtype, device=device),
+        )
+        self.frequency_embedding_size = frequency_embedding_size
+
+    @staticmethod
+    def timestep_embedding(t, dim, max_period=10000):
+        half = dim // 2
+        freqs = torch.exp(-np.log(max_period) * torch.arange(start=0, end=half, dtype=torch.float32) / half).to(device=t.device)
+        args = t[:, None].float() * freqs[None]
+        embedding = torch.cat([torch.cos(args), torch.sin(args)], dim=-1)
+        if dim % 2:
+            embedding = torch.cat([embedding, torch.zeros_like(embedding[:, :1])], dim=-1)
+        return embedding
+
+    def forward(self, t):
+        emb = self.timestep_embedding(t, self.frequency_embedding_size)
+        return self.mlp(emb.to(self.mlp[0].weight.dtype))
+
+
+class LatentSeqMMFlowModel(nn.Module):
+    def __init__(self, image_model=None, q_token_length=8192, in_channels=16, model_channels=1024,
+                 cond_channels=1280, out_channels=16, num_blocks=24, num_refiner_blocks=2,
+                 num_heads=None, num_head_channels=64, cam_channels=5, cond2_channels=128,
+                 mlp_ratio=4, share_mod=True, qk_rms_norm=True,
+                 dtype=None, device=None, operations=None, **kwargs):
+        super().__init__()
+        self.dtype = dtype
+        self.q_token_length = q_token_length
+        self.in_channels = in_channels
+        self.cam_channels = cam_channels
+        self.model_channels = model_channels
+        self.cond_channels = cond_channels
+        self.cond2_channels = cond2_channels
+        self.out_channels = out_channels
+        self.num_blocks = num_blocks
+        self.num_refiner_blocks = num_refiner_blocks
+        self.num_heads = num_heads or model_channels // num_head_channels
+        self.mlp_ratio = mlp_ratio
+        self.share_mod = share_mod
+        self.qk_rms_norm = qk_rms_norm
+
+        factory_kwargs = dict(dtype=dtype, device=device)
+        op_kwargs = dict(operations=operations, **factory_kwargs)
+
+        self.t_embedder = TimestepEmbedder(model_channels, **op_kwargs)
+        if share_mod:
+            self.adaLN_modulation = nn.Sequential(nn.SiLU(), operations.Linear(model_channels, 6 * model_channels, bias=True, **factory_kwargs))
+
+        self.input_layer = operations.Linear(in_channels, model_channels, **factory_kwargs)
+        self.cond_embedder = operations.Linear(cond_channels, model_channels, **factory_kwargs)
+        self.cond_embedder2 = operations.Linear(cond2_channels, model_channels, **factory_kwargs) if cond2_channels is not None else None
+
+        # Fixed Sobol (low-discrepancy) 3D anchor positions for the latent tokens, used as positional encoding.
+        # The embedder is parameter-free and the anchors are fixed, precompute once.
+        sobol_seq = torch.quasirandom.SobolEngine(dimension=3, scramble=True, seed=123).draw(q_token_length)
+        pos_emb = PcdAbsolutePositionEmbedder(model_channels)(sobol_seq.unsqueeze(0))
+        self.register_buffer("pos_emb", pos_emb, persistent=False)
+
+        # RePo3DRotaryEmbedding layers for the refiner and main blocks
+        repo_kwargs = dict(num_heads=self.num_heads, head_dim=num_head_channels, **op_kwargs)
+        self.noise_repo_layers = nn.ModuleList(
+            [RePo3DRotaryEmbedding(model_channels, **repo_kwargs) for _ in range(num_refiner_blocks)])
+        self.context_repo_layers = nn.ModuleList(
+            [RePo3DRotaryEmbedding(model_channels, **repo_kwargs) for _ in range(num_refiner_blocks)])
+        self.repo_layers = nn.ModuleList(
+            [RePo3DRotaryEmbedding(model_channels, **repo_kwargs) for _ in range(num_blocks)])
+
+        # Refiner blocks
+        block_kwargs = dict(num_heads=self.num_heads, mlp_ratio=self.mlp_ratio, use_rope=True, qk_rms_norm=self.qk_rms_norm, **op_kwargs)
+        self.noise_refiner = nn.ModuleList(
+            [UnifiedTransformerBlock(model_channels, modulation=True, share_mod=self.share_mod, **block_kwargs) for _ in range(num_refiner_blocks)])
+        self.context_refiner = nn.ModuleList(
+            [UnifiedTransformerBlock(model_channels, modulation=False, **block_kwargs) for _ in range(num_refiner_blocks)])
+
+        self.cam_refiner = MLP(self.cam_channels, model_channels, model_channels, **op_kwargs)
+
+        self.blocks = nn.ModuleList(
+            [UnifiedTransformerBlock(model_channels, modulation=True, share_mod=self.share_mod, **block_kwargs) for _ in range(num_blocks)])
+
+        self.shift_table = nn.Parameter(torch.empty(1, 2, model_channels, **factory_kwargs))
+        self.out_layer = operations.Linear(model_channels, out_channels, **factory_kwargs)
+        self.cam_out_layer = operations.Linear(model_channels, cam_channels, **factory_kwargs)
+
+    def forward(self, x, t, context=None, ref_latents=None, transformer_options={}, **kwargs):
+        return comfy.patcher_extension.WrapperExecutor.new_class_executor(
+            self._forward,
+            self,
+            comfy.patcher_extension.get_all_wrappers(comfy.patcher_extension.WrappersMP.DIFFUSION_MODEL, transformer_options)
+        ).execute(x, t, context, ref_latents, transformer_options, **kwargs)
+
+    def _forward(self, x, t, context=None, ref_latents=None, transformer_options={}, **kwargs):
+        # x is the unpacked nested latent: [latent (B,8192,in_channels), camera (B,1,cam_channels)].
+        # context == feature1.
+        z, camera = x[0], x[1]
+        feat1 = context
+
+        h_x = self.input_layer(z)
+        h_cond = self.cond_embedder(feat1)
+        if ref_latents is not None and self.cond_embedder2 is not None:
+            # Flatten the Flux2 VAE latent (B,128,h,w) to a token sequence and front-pad to feat1's length
+            # (the pad count = feat1's prefix tokens: DINOv3 cls + registers), then add to the context.
+            feat2 = ref_latents[0].flatten(2).transpose(1, 2)
+            feat2 = F.pad(feat2, (0, 0, feat1.shape[1] - feat2.shape[1], 0))
+            h_cond = h_cond + self.cond_embedder2(feat2.to(h_cond.dtype))
+        t_emb = self.t_embedder(t)
+        t_mod = self.adaLN_modulation(t_emb) if self.share_mod else t_emb
+
+        h_x = h_x + self.pos_emb.to(z)
+
+        for i, block in enumerate(self.noise_refiner):
+            h_x = block(h_x, mod=t_mod, rotary_emb=self.noise_repo_layers[i](h_x), transformer_options=transformer_options)
+
+        for i, block in enumerate(self.context_refiner):
+            h_cond = block(h_cond, mod=None, rotary_emb=self.context_repo_layers[i](h_cond), transformer_options=transformer_options)
+
+        cam = camera.to(z)
+        h_cam = self.cam_refiner(cam)
+        h = torch.cat([h_x, h_cond, h_cam], dim=1)
+
+        for i, block in enumerate(self.blocks):
+            h = block(h, mod=t_mod, rotary_emb=self.repo_layers[i](h), transformer_options=transformer_options)
+
+        h_x = F.layer_norm(h[:, :z.shape[1]].float(), h.shape[-1:]).to(z)
+        h_cam = F.layer_norm(h[:, -cam.shape[1]:].float(), h.shape[-1:]).to(z)
+
+        shift, scale = (comfy.model_management.cast_to(self.shift_table, t_emb.dtype, t_emb.device) + t_emb.unsqueeze(1)).chunk(2, dim=1)
+        scale = 1 + scale
+        h_x = torch.addcmul(shift, h_x, scale)
+        h_cam = torch.addcmul(shift, h_cam, scale)
+
+        return self.out_layer(h_x), self.cam_out_layer(h_cam)

comfy/ldm/triposplat/preview.py

@@ -0,0 +1,91 @@
+# Live preview for TripoSplat: decode an x0 estimate into a coarse gaussian splat and render it with a perspective orbit camera.
+import numpy as np
+from PIL import Image
+
+_C0 = 0.28209479177387814
+_LATENT_TOKENS = 8192  # q_token_length
+_LATENT_CH = 16        # in_channels
+_OBJECT_TO_VIEWER = np.array([[1, 0, 0], [0, 0, -1], [0, 1, 0]], np.float32)  # object frame -> viewer Y-up frame
+
+
+def _view_matrix(yaw_deg, pitch_deg):
+    y, p = np.radians(yaw_deg), np.radians(pitch_deg)
+    Ry = np.array([[np.cos(y), 0, np.sin(y)], [0, 1, 0], [-np.sin(y), 0, np.cos(y)]], np.float32)
+    Rx = np.array([[1, 0, 0], [0, np.cos(p), -np.sin(p)], [0, np.sin(p), np.cos(p)]], np.float32)
+    return Rx @ Ry
+
+
+def render_splat(xyz, rgb, scale, opacity=None, yaw=35.0, pitch=30.0, size=320, min_px=2, gain=1.0,
+                 max_px=9, min_opacity=0.0, fov=35.0, dist=2.2):
+    # Project gaussian centers with a perspective camera and paint each as a filled disk whose screen
+    # radius follows the gaussian's world-space scale, composited with a nearest-wins z-buffer.
+    # gain scales the footprint (≈ std spanned), `min_px`/`max_px` clamp the on-screen radius.
+
+    pts = xyz.astype(np.float32) @ _OBJECT_TO_VIEWER.T
+    v = pts @ _view_matrix(yaw, pitch).T
+    zc = v[:, 2] + dist
+    keep = zc > 1e-2
+    if opacity is not None and min_opacity > 0.0:  # culls gaussians with very low opacity
+        keep = keep & (opacity > min_opacity)
+    v, zc, scale = v[keep], zc[keep], scale[keep]
+    col = (np.clip(rgb, 0, 1)[:, :3] * 255).astype(np.uint8)[keep]
+    if v.shape[0] == 0:
+        return Image.fromarray(np.zeros((size, size, 3), np.uint8))
+    f = (size / 2) / np.tan(np.radians(fov) / 2)
+    cx = size / 2 + f * v[:, 0] / zc
+    cy = size / 2 + f * v[:, 1] / zc
+    radius = np.clip(np.round(f * scale / zc * gain), min_px, max_px).astype(np.int32)
+
+    # Expand each splat to its disk pixels, bucketed by integer radius so it stays vectorized.
+    px, py, pz, pc = [], [], [], []
+    for r in range(int(radius.min()), int(radius.max()) + 1):
+        m = radius == r
+        if not m.any():
+            continue
+        dy, dx = np.mgrid[-r:r + 1, -r:r + 1]
+        disk = (dx * dx + dy * dy) <= r * r
+        ox, oy = dx[disk], dy[disk]
+        px.append((cx[m, None] + ox).ravel())
+        py.append((cy[m, None] + oy).ravel())
+        pz.append(np.repeat(zc[m], ox.size))
+        pc.append(np.repeat(col[m], ox.size, axis=0))
+    px, py = np.concatenate(px), np.concatenate(py)
+    pz, pc = np.concatenate(pz), np.concatenate(pc)
+    xi = np.clip(px, 0, size - 1).astype(np.int64)
+    yi = np.clip(py, 0, size - 1).astype(np.int64)
+
+    # Nearest-wins z-buffer: pack (quantized depth, source index), per-pixel min picks the closest
+    # splat, then decode the winning index back to its color.
+    pid = yi * size + xi
+    q = np.clip((pz * 1024.0).astype(np.int64), 0, (1 << 20) - 1)  # near = small
+    key = (q << 32) | np.arange(pid.size, dtype=np.int64)
+    buf = np.full(size * size, 1 << 62, np.int64)
+    np.minimum.at(buf, pid, key)
+    img = np.zeros((size * size, 3), np.uint8)
+    hit = buf < (1 << 62)
+    img[hit] = pc[buf[hit] & 0xFFFFFFFF]
+    return Image.fromarray(img.reshape(size, size, 3))
+
+
+def _extract_latent(x0):
+    # x0 from the sampler callback is the nested latent packed to (B, 1, TOKENS*CH + 1*5);
+    # the plain single-latent case is (B, TOKENS, CH). Return the (B, TOKENS, CH) latent stream.
+    if x0.ndim == 3 and x0.shape[1] == _LATENT_TOKENS and x0.shape[2] == _LATENT_CH:
+        return x0
+    flat = x0.reshape(x0.shape[0], -1)
+    return flat[:, :_LATENT_TOKENS * _LATENT_CH].reshape(x0.shape[0], _LATENT_TOKENS, _LATENT_CH)
+
+
+def decode_x0_to_image(decoder, x0, cfg):
+    # Decode x0 at a coarse octree level / few gaussians and render a preview image.
+    latent = _extract_latent(x0)
+    fsm = decoder.first_stage_model
+    gaussian = fsm.decode(latent.to(decoder.device, decoder.vae_dtype),
+                          num_gaussians=cfg.get("gaussians", 16384), level=cfg.get("level", 5))[0]
+    xyz = gaussian.get_xyz.float().cpu().numpy()
+    rgb = gaussian._features_dc.float().cpu().numpy()[:, 0, :] * _C0 + 0.5
+    scale = gaussian.get_scaling.float().cpu().numpy().max(axis=1)  # per-splat world radius (largest axis)
+    opacity = gaussian.get_opacity.float().cpu().numpy()[:, 0]
+    return render_splat(xyz, rgb, scale, opacity=opacity, yaw=cfg.get("yaw", 35.0), pitch=cfg.get("pitch", 30.0),
+                        size=cfg.get("size", 320), min_px=1, gain=1.0, max_px=cfg.get("point_size", 3),
+                        min_opacity=0.01)

comfy/ldm/triposplat/vae.py

@@ -0,0 +1,382 @@
+# TripoSplat gaussian decoder ("VAE"): an octree probability decoder picks point coords, then an
+# elastic-gaussian decoder predicts per-point gaussian params. OctreeGaussianDecoder.decode() returns
+# a Gaussian. The octree sampler uses the global torch RNG (no generator) like upstream, so seed it for repeatable decodes.
+import numpy as np
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+import comfy.model_management
+import comfy.ops
+from .gaussian import build_gaussian_models
+from .model import MultiHeadRMSNorm, MLP, PcdAbsolutePositionEmbedder, attention
+
+
+# Quasi-random sampling utilities (pure functions, dtype/device-agnostic)
+
+PRIMES = [2, 3, 5, 7, 11, 13, 17, 19, 23, 29, 31, 37, 41, 43, 47, 53]
+
+
+def radical_inverse(base, n):
+    val = 0
+    inv_base = 1.0 / base
+    inv_base_n = inv_base
+    while n > 0:
+        digit = n % base
+        val += digit * inv_base_n
+        n //= base
+        inv_base_n *= inv_base
+    return val
+
+
+def halton_sequence(dim, n):
+    return [radical_inverse(PRIMES[i], n) for i in range(dim)]
+
+
+def hammersley_sequence(dim, n, num_samples):
+    return [n / num_samples] + halton_sequence(dim - 1, n)
+
+
+def sample_probs(probs, counts, generator=None):
+    # Systematic resampling: distribute counts[r] draws across the P bins of row r
+    batch_shape = counts.shape
+    R = counts.numel()
+    P = probs.size(-1)
+    device = probs.device
+    probs = probs.reshape(R, P).to(torch.float32).clamp_min(0)
+    counts = counts.reshape(R).to(device=device, dtype=torch.long)
+
+    row_sums = probs.sum(1, keepdim=True)
+    probs = torch.where(row_sums == 0, probs.new_tensor(1.0 / P), probs / row_sums.clamp_min(1))
+    cdf = probs.cumsum(dim=1).clamp(max=1.0 - 1e-12)
+
+    Nmax = int(counts.max())
+    if Nmax == 0:
+        return counts.new_zeros(*batch_shape, P)
+    cnt = counts.clamp_min(1).float().unsqueeze(1)                              # (R, 1)
+    grid = torch.arange(Nmax, device=device, dtype=torch.float32).unsqueeze(0)  # (1, Nmax)
+    u = (torch.rand(R, 1, generator=generator).to(device) + grid) / cnt         # (R, Nmax) systematic samples (CPU-seeded)
+    idx = torch.searchsorted(cdf, u.clamp(max=1.0 - 1e-12)).clamp_max(P - 1)
+    weight = (grid < counts.unsqueeze(1)).to(cdf.dtype)                         # mask out j >= counts[r]
+    out = torch.zeros(R, P, dtype=torch.float32, device=device)
+    out.scatter_add_(1, idx, weight)
+    return out.to(torch.long).view(*batch_shape, P)
+
+
+class MultiHeadAttention(nn.Module):
+    def __init__(self, channels, num_heads, ctx_channels=None, type="self", qkv_bias=True, qk_rms_norm=False,
+                 dtype=None, device=None, operations=None):
+        super().__init__()
+        assert channels % num_heads == 0
+        self.channels = channels
+        self.head_dim = channels // num_heads
+        self.ctx_channels = ctx_channels if ctx_channels is not None else channels
+        self.num_heads = num_heads
+        self._type = type
+        self.qk_rms_norm = qk_rms_norm
+        if self._type == "self":
+            self.to_qkv = operations.Linear(channels, channels * 3, bias=qkv_bias, dtype=dtype, device=device)
+        else:
+            self.to_q = operations.Linear(channels, channels, bias=qkv_bias, dtype=dtype, device=device)
+            self.to_kv = operations.Linear(self.ctx_channels, channels * 2, bias=qkv_bias, dtype=dtype, device=device)
+        if self.qk_rms_norm:
+            self.q_rms_norm = MultiHeadRMSNorm(self.head_dim, num_heads, dtype=dtype, device=device)
+            self.k_rms_norm = MultiHeadRMSNorm(self.head_dim, num_heads, dtype=dtype, device=device)
+        self.to_out = operations.Linear(channels, channels, dtype=dtype, device=device)
+
+    def forward(self, x, context=None):
+        B, L, C = x.shape
+        if self._type == "self":
+            q, k, v = self.to_qkv(x).reshape(B, L, 3, self.num_heads, -1).unbind(dim=2)
+        else:
+            Lkv = context.shape[1]
+            q = self.to_q(x).reshape(B, L, self.num_heads, -1)
+            k, v = self.to_kv(context).reshape(B, Lkv, 2, self.num_heads, -1).unbind(dim=2)
+        if self.qk_rms_norm:
+            q = self.q_rms_norm(q)
+            k = self.k_rms_norm(k)
+        h = attention(q, k, v)
+        return self.to_out(h.reshape(B, L, -1))
+
+
+# Octree probability decoder
+
+class LevelEmbedder(nn.Module):
+    def __init__(self, hidden_size, frequency_embedding_size=256, max_period=1024,
+                 dtype=None, device=None, operations=None):
+        super().__init__()
+        self.mlp = nn.Sequential(
+            operations.Linear(frequency_embedding_size, hidden_size, bias=True, dtype=dtype, device=device),
+            nn.SiLU(),
+            operations.Linear(hidden_size, hidden_size, bias=True, dtype=dtype, device=device),
+        )
+        self.frequency_embedding_size = frequency_embedding_size
+        self.max_period = max_period
+
+    @staticmethod
+    def level_embedding(t, dim, max_period=1024):
+        half = dim // 2
+        freqs = torch.exp(-np.log(max_period) * torch.arange(start=0, end=half, dtype=torch.float32) / half).to(device=t.device)
+        args = t[:, None].float() * freqs[None] * 2 * torch.pi
+        embedding = torch.cat([torch.cos(args), torch.sin(args)], dim=-1)
+        if dim % 2:
+            embedding = torch.cat([embedding, torch.zeros_like(embedding[:, :1])], dim=-1)
+        return embedding
+
+    def forward(self, t):
+        emb = self.level_embedding(t, self.frequency_embedding_size, self.max_period)
+        return self.mlp(emb.to(self.mlp[0].weight.dtype))
+
+
+class ModulatedTransformerCrossOnlyBlock(nn.Module):
+    def __init__(self, channels, ctx_channels, num_heads, mlp_ratio=4.0, share_mod=False,
+                 qk_rms_norm_cross=True, qkv_bias=True, dtype=None, device=None, operations=None):
+        super().__init__()
+        self.share_mod = share_mod
+        self.norm1 = operations.LayerNorm(channels, elementwise_affine=False, eps=1e-6, dtype=dtype, device=device)
+        self.norm2 = operations.LayerNorm(channels, elementwise_affine=False, eps=1e-6, dtype=dtype, device=device)
+        self.cross_attn = MultiHeadAttention(channels, ctx_channels=ctx_channels, num_heads=num_heads,
+                                             type="cross", qkv_bias=qkv_bias,
+                                             qk_rms_norm=qk_rms_norm_cross, dtype=dtype, device=device, operations=operations)
+        self.mlp = MLP(channels, int(channels * mlp_ratio), channels, dtype=dtype, device=device, operations=operations)
+        if not share_mod:
+            self.adaLN_modulation = nn.Sequential(
+                nn.SiLU(), operations.Linear(channels, 6 * channels, bias=True, dtype=dtype, device=device))
+
+    def forward(self, x, mod, context):
+        if self.share_mod:
+            shift_msa, scale_msa, gate_msa, shift_mlp, scale_mlp, gate_mlp = mod.chunk(6, dim=1)
+        else:
+            shift_msa, scale_msa, gate_msa, shift_mlp, scale_mlp, gate_mlp = self.adaLN_modulation(mod).chunk(6, dim=1)
+        h = torch.addcmul(shift_msa.unsqueeze(1), self.norm1(x), 1 + scale_msa.unsqueeze(1))
+        x = torch.addcmul(x, self.cross_attn(h, context), gate_msa.unsqueeze(1))
+        h = torch.addcmul(shift_mlp.unsqueeze(1), self.norm2(x), 1 + scale_mlp.unsqueeze(1))
+        x = torch.addcmul(x, self.mlp(h), gate_mlp.unsqueeze(1))
+        return x
+
+
+class OctreeProbabilityFixedlenDecoder(nn.Module):
+    # Cross-attention transformer over octree coords -> per-node 8-way child occupancy logits.
+    def __init__(self, model_channels=1024, cond_channels=16, num_blocks=4, num_heads=16,
+                 num_head_channels=64, mlp_ratio=4.0, share_mod=True,
+                 qk_rms_norm_cross=True, dtype=None, device=None, operations=None):
+        super().__init__()
+        self.model_channels = model_channels
+        self.cond_channels = cond_channels
+        self.num_blocks = num_blocks
+        self.num_heads = num_heads or model_channels // num_head_channels
+        self.mlp_ratio = mlp_ratio
+        self.share_mod = share_mod
+        self.qk_rms_norm_cross = qk_rms_norm_cross
+        self.input_layer = operations.Linear(model_channels, model_channels, dtype=dtype, device=device)
+        self.l_embedder = LevelEmbedder(model_channels, dtype=dtype, device=device, operations=operations)
+        if share_mod:
+            self.adaLN_modulation = nn.Sequential(
+                nn.SiLU(), operations.Linear(model_channels, 6 * model_channels, bias=True, dtype=dtype, device=device))
+        if cond_channels is not None:
+            self.blocks = nn.ModuleList([
+                ModulatedTransformerCrossOnlyBlock(
+                    model_channels, ctx_channels=cond_channels, num_heads=self.num_heads,
+                    mlp_ratio=self.mlp_ratio, qk_rms_norm_cross=self.qk_rms_norm_cross,
+                    share_mod=self.share_mod, dtype=dtype, device=device, operations=operations)
+                for _ in range(num_blocks)
+            ])
+        self.out_proj = operations.Linear(model_channels, 8, dtype=dtype, device=device)
+        self.in_proj = operations.Linear(3, model_channels, dtype=dtype, device=device)
+        self.pos_embedder = PcdAbsolutePositionEmbedder(channels=model_channels, in_channels=3, max_res=10, schedule="log2")
+
+    def forward(self, x, l, cond):
+        d = next(self.parameters()).dtype
+        B, L, _ = x.shape
+        h = self.in_proj(x.to(d)) + self.pos_embedder(x.reshape(-1, 3)).reshape(B, L, -1).to(d)
+        h = self.input_layer(h)
+        l_emb = self.l_embedder(l)
+        if self.share_mod:
+            l_emb = self.adaLN_modulation(l_emb)
+        cond = cond.to(d)
+        for block in self.blocks:
+            h = block(h, l_emb, cond)
+        h = F.layer_norm(h.float(), h.shape[-1:]).to(d)
+        logits = self.out_proj(h)
+        return {"logits": logits, "probs": torch.softmax(logits, dim=-1)}
+
+    @staticmethod
+    def sample(model, cond, num_points, level, temperature=1.0, generator=None):
+        B = cond.shape[0]
+        device = cond.device
+        child_offset = torch.tensor([[i, j, k] for k in [0, 1] for j in [0, 1] for i in [0, 1]],
+                                    dtype=torch.long, device=device)
+        prev_coords_int = torch.zeros(B, 1, 3, dtype=torch.long, device=device)
+        prev_counts = torch.full((B, 1), num_points, dtype=torch.long, device=device)
+        prev_log_probs = torch.zeros(B, 1, dtype=torch.float32, device=device)
+        batch_indices_range = torch.arange(B, device=device).unsqueeze(1)
+
+        for lv in range(1, level + 1):
+            res_p = 1 << (lv - 1)
+            res = 1 << lv
+            parent_coords_norm = (prev_coords_int.to(torch.float32) + 0.5) / res_p
+            res_tensor = torch.full((B,), res, dtype=torch.long, device=device)
+            pred_logits = model(parent_coords_norm, res_tensor, cond)["logits"] / temperature
+            pred_probs = torch.softmax(pred_logits, dim=-1)
+            pred_log_probs = torch.log_softmax(pred_logits, dim=-1)
+            sampled = sample_probs(pred_probs, prev_counts, generator=generator).flatten(1, 2)
+            pred_log_probs = pred_log_probs.flatten(1, 2)
+            prev_log_probs_expanded = prev_log_probs.repeat_interleave(8, dim=1)
+            child_coords_int = (prev_coords_int[:, :, None, :] * 2 + child_offset[None, None, :, :]).flatten(1, 2)
+            mask = sampled > 0
+            max_valid = mask.sum(dim=1).max().item()
+            scatter_indices = mask.cumsum(dim=1) - 1
+            valid_scatter_indices = scatter_indices[mask]
+            valid_batch_indices = batch_indices_range.expand_as(mask)[mask]
+            next_prev_coords_int = torch.zeros(B, max_valid, 3, dtype=child_coords_int.dtype, device=device)
+            next_prev_coords_int[valid_batch_indices, valid_scatter_indices] = child_coords_int[mask]
+            next_prev_counts = torch.zeros(B, max_valid, dtype=sampled.dtype, device=device)
+            next_prev_counts[valid_batch_indices, valid_scatter_indices] = sampled[mask]
+            next_prev_log_probs = torch.zeros(B, max_valid, dtype=prev_log_probs.dtype, device=device)
+            next_prev_log_probs[valid_batch_indices, valid_scatter_indices] = (prev_log_probs_expanded + pred_log_probs)[mask]
+            prev_coords_int = next_prev_coords_int
+            prev_counts = next_prev_counts
+            prev_log_probs = next_prev_log_probs
+
+        res = 1 << level
+        prev_log_probs = torch.repeat_interleave(prev_log_probs.flatten(0, 1), prev_counts.flatten(0, 1), dim=0).reshape(B, num_points)
+        coords_int = torch.repeat_interleave(prev_coords_int.flatten(0, 1), prev_counts.flatten(0, 1), dim=0).reshape(B, num_points, -1)
+        rand = torch.rand(coords_int.shape, dtype=torch.float32, generator=generator).to(device)
+        coords_norm = (coords_int.to(torch.float32) + rand) / res
+        return {"points": coords_norm, "log_probs": prev_log_probs}
+
+
+# Elastic gaussian decoder
+
+class TransformerCrossBlock(nn.Module):
+    def __init__(self, channels, ctx_channels, num_heads, mlp_ratio=4.0,
+                 qk_rms_norm=True, qk_rms_norm_cross=True, qkv_bias=True,
+                 dtype=None, device=None, operations=None):
+        super().__init__()
+        self.norm1 = operations.LayerNorm(channels, elementwise_affine=False, eps=1e-6, dtype=dtype, device=device)
+        self.norm2 = operations.LayerNorm(channels, elementwise_affine=True, eps=1e-6, dtype=dtype, device=device)
+        self.norm3 = operations.LayerNorm(channels, elementwise_affine=False, eps=1e-6, dtype=dtype, device=device)
+        self.self_attn = MultiHeadAttention(channels, num_heads=num_heads, type="self", qkv_bias=qkv_bias,
+                                            qk_rms_norm=qk_rms_norm, dtype=dtype, device=device, operations=operations)
+        self.cross_attn = MultiHeadAttention(channels, ctx_channels=ctx_channels, num_heads=num_heads, type="cross",
+                                             qkv_bias=qkv_bias, qk_rms_norm=qk_rms_norm_cross, dtype=dtype, device=device, operations=operations)
+        self.mlp = MLP(channels, int(channels * mlp_ratio), channels, dtype=dtype, device=device, operations=operations)
+
+    def forward(self, x, context):
+        x = x + self.self_attn(self.norm1(x))
+        x = x + self.cross_attn(self.norm2(x), context)
+        x = x + self.mlp(self.norm3(x))
+        return x
+
+
+class ElasticGaussianFixedlenDecoder(nn.Module):
+    # Cross-attention transformer over sampled octree points -> per-point gaussian params.
+    def __init__(self, in_channels=3, model_channels=1024, cond_channels=16, num_blocks=16, num_heads=16,
+                 num_head_channels=64, mlp_ratio=4.0, *, representation_config=None,
+                 qk_rms_norm=True, qk_rms_norm_cross=True, dtype=None, device=None, operations=None):
+        super().__init__()
+        self.rep_config = representation_config or dict(
+            lr=dict(_xyz=1.0, _features_dc=1.0, _opacity=1.0, _scaling=1.0, _rotation=0.1),
+            perturb_offset=True, perturbe_size=1.5, offset_scale=0.05, num_gaussians=32,
+            filter_kernel_size_3d=0.0009, scaling_bias=0.004, opacity_bias=0.1,
+            scaling_activation="softplus",
+        )
+        self.out_channels = self._calc_layout()
+        self.model_channels = model_channels
+        self.cond_channels = cond_channels
+        self.num_blocks = num_blocks
+        self.num_heads = num_heads or model_channels // num_head_channels
+        self.mlp_ratio = mlp_ratio
+        self.input_layer = operations.Linear(model_channels, model_channels, dtype=dtype, device=device)
+        if cond_channels is not None:
+            self.blocks = nn.ModuleList([
+                TransformerCrossBlock(model_channels, ctx_channels=cond_channels,
+                                      num_heads=self.num_heads, mlp_ratio=self.mlp_ratio,
+                                      qk_rms_norm=qk_rms_norm, qk_rms_norm_cross=qk_rms_norm_cross,
+                                      dtype=dtype, device=device, operations=operations)
+                for _ in range(num_blocks)
+            ])
+        self.in_proj = operations.Linear(in_channels, model_channels, dtype=dtype, device=device)
+        self.pos_embedder = PcdAbsolutePositionEmbedder(channels=model_channels, in_channels=3, max_res=10, schedule="log2")
+        self.out_proj = operations.Linear(model_channels, self.out_channels, dtype=dtype, device=device)
+        self._build_perturbation()
+
+    def _calc_layout(self):
+        ng = self.rep_config['num_gaussians']
+        self.layout = {
+            '_xyz':         {'shape': (ng, 3),    'size': ng * 3},
+            '_features_dc': {'shape': (ng, 1, 3), 'size': ng * 3},
+            '_scaling':     {'shape': (ng, 3),    'size': ng * 3},
+            '_rotation':    {'shape': (ng, 4),    'size': ng * 4},
+            '_opacity':     {'shape': (ng, 1),    'size': ng},
+        }
+        self.layout['_offset_scale'] = {'shape': (ng, 1), 'size': ng}
+        start = 0
+        for k, v in self.layout.items():
+            v['range'] = (start, start + v['size'])
+            start += v['size']
+        return start
+
+    def _build_perturbation(self):
+        ng = self.rep_config['num_gaussians']
+        perturbation = torch.tensor([hammersley_sequence(3, i, ng) for i in range(ng)]).float()
+        perturbation = torch.atanh((perturbation * 2 - 1) / self.rep_config['perturbe_size'])
+        self.register_buffer('points_offset_perturbation', perturbation)
+        base = torch.tensor(self.rep_config['offset_scale'])
+        self.register_buffer('base_offset_scale', torch.log(torch.exp(base) - 1.0))
+
+    def _get_offset(self, h):
+        B = h.shape[0]
+        r = self.layout['_offset_scale']['range']
+        _offset_scale = F.softplus(
+            h[:, :, r[0]:r[1]].reshape(B, -1, *self.layout['_offset_scale']['shape'])
+            + comfy.model_management.cast_to(self.base_offset_scale, h.dtype, h.device))
+
+        r = self.layout['_xyz']['range']
+        offset = h[:, :, r[0]:r[1]].reshape(B, -1, *self.layout['_xyz']['shape'])
+        offset = offset * self.rep_config['lr']['_xyz']
+        if self.rep_config['perturb_offset']:
+            offset = offset + comfy.model_management.cast_to(self.points_offset_perturbation, offset.dtype, offset.device)
+        offset = torch.tanh(offset) * 0.5 * self.rep_config['perturbe_size']
+        offset = offset * _offset_scale
+        return offset
+
+    def forward(self, x=None, cond=None):
+        pcd = x["points"]
+        d = next(self.parameters()).dtype
+        B, L, _ = pcd.shape
+        h = self.in_proj(pcd.to(d)) + self.pos_embedder(pcd.reshape(-1, 3)).reshape(B, L, -1).to(d)
+        h = self.input_layer(h)
+        cond = cond.to(d)
+        for block in self.blocks:
+            h = block(h, cond)
+        h = F.layer_norm(h.float(), h.shape[-1:]).to(h.dtype)
+        return {"features": self.out_proj(h)}
+
+
+# Combined octree gaussian decoder (comfy first-stage model)
+
+class OctreeGaussianDecoder(nn.Module):
+    _MAX_VOXEL_LEVEL = 8
+
+    def __init__(self, dtype=None, device=None, operations=None):
+        super().__init__()
+        if operations is None:
+            operations = comfy.ops.disable_weight_init
+        self.octree = OctreeProbabilityFixedlenDecoder(dtype=dtype, device=device, operations=operations)
+        self.gs = ElasticGaussianFixedlenDecoder(dtype=dtype, device=device, operations=operations)
+
+    @property
+    def gaussians_per_point(self) -> int:
+        return self.gs.rep_config['num_gaussians']
+
+    def decode(self, latent: torch.Tensor, num_gaussians: int, level: int = None, generator=None):
+        # level defaults to the full octree depth, a lower level is cheaper (coarser) for live previews.
+        # generator (a CPU torch.Generator) makes the octree sampling reproducible without touching global RNG.
+        level = self._MAX_VOXEL_LEVEL if level is None else level
+        num_decoder_tokens = max(1, num_gaussians // self.gaussians_per_point)
+        points_pred = OctreeProbabilityFixedlenDecoder.sample(
+            self.octree, latent, num_points=num_decoder_tokens, level=level, temperature=1.0, generator=generator,
+        )
+        pred = self.gs(x=points_pred, cond=latent)
+        return build_gaussian_models(self.gs, points_pred, pred)  # one GaussianModel per batch item

comfy/ldm/wan/model.py

@@ -8,7 +8,7 @@ from einops import rearrange
 
 from comfy.ldm.modules.attention import optimized_attention
 from comfy.ldm.flux.layers import EmbedND
-from comfy.ldm.flux.math import apply_rope1
+from comfy.ldm.flux.math import apply_rope1, rope
 import comfy.ldm.common_dit
 import comfy.model_management
 import comfy.patcher_extension
@@ -570,6 +570,14 @@ class WanModel(torch.nn.Module):
                 full_ref = self.ref_conv(full_ref).flatten(2).transpose(1, 2)
                 x = torch.concat((full_ref, x), dim=1)
 
+        # In-context reference (Bernini)
+        context_latents = kwargs.get("context_latents", None)
+        main_len = x.shape[1]
+        if context_latents is not None:
+            for lat in context_latents:
+                cl = self.patch_embedding(lat.float().to(x.device)).to(x.dtype).flatten(2).transpose(1, 2)
+                x = torch.cat([x, cl], dim=1)
+
         # context
         context = self.text_embedding(context)
 
@@ -599,6 +607,9 @@ class WanModel(torch.nn.Module):
         # head
         x = self.head(x, e)
 
+        if context_latents is not None:
+            x = x[:, :main_len]
+
         if full_ref is not None:
             x = x[:, full_ref.shape[1]:]
 
@@ -606,7 +617,7 @@ class WanModel(torch.nn.Module):
         x = self.unpatchify(x, grid_sizes)
         return x
 
-    def rope_encode(self, t, h, w, t_start=0, steps_t=None, steps_h=None, steps_w=None, device=None, dtype=None, transformer_options={}):
+    def rope_encode(self, t, h, w, t_start=0, steps_t=None, steps_h=None, steps_w=None, device=None, dtype=None, transformer_options={}, source_id=0):
         patch_size = self.patch_size
         t_len = ((t + (patch_size[0] // 2)) // patch_size[0])
         h_len = ((h + (patch_size[1] // 2)) // patch_size[1])
@@ -638,6 +649,13 @@ class WanModel(torch.nn.Module):
         img_ids = img_ids.reshape(1, -1, img_ids.shape[-1])
 
         freqs = self.rope_embedder(img_ids).movedim(1, 2)
+
+        # In-context reference: a non-zero source_id composes an extra rotation into the spatial rope
+        if source_id:
+            d = self.dim // self.num_heads
+            pos = torch.tensor([[float(source_id)]], device=freqs.device, dtype=torch.float32)
+            id_rot = rope(pos, d, self.rope_embedder.theta).reshape(1, 1, 1, d // 2, 2, 2).to(freqs.dtype)
+            freqs = torch.einsum('...ij,...jk->...ik', freqs, id_rot)
         return freqs
 
     def forward(self, x, timestep, context, clip_fea=None, time_dim_concat=None, transformer_options={}, **kwargs):
@@ -661,6 +679,15 @@ class WanModel(torch.nn.Module):
             t_len += 1
 
         freqs = self.rope_encode(t_len, h, w, device=x.device, dtype=x.dtype, transformer_options=transformer_options)
+
+        # In-context reference: one rope block per stream, each with it's own source_id (1, 2, ...) to distinguish from the target (id 0).
+        context_latents = kwargs.get("context_latents", None)
+        if context_latents is not None:
+            context_latents = [comfy.ldm.common_dit.pad_to_patch_size(lat, self.patch_size) for lat in context_latents]
+            for i, lat in enumerate(context_latents):
+                freqs = torch.cat([freqs, self.rope_encode(lat.shape[-3], lat.shape[-2], lat.shape[-1], device=x.device, dtype=x.dtype, transformer_options=transformer_options, source_id=i + 1)], dim=1)
+            kwargs = {**kwargs, "context_latents": context_latents}
+
         return self.forward_orig(x, timestep, context, clip_fea=clip_fea, freqs=freqs, transformer_options=transformer_options, **kwargs)[:, :, :t, :h, :w]
 
     def unpatchify(self, x, grid_sizes):
@@ -1631,13 +1658,15 @@ class SCAILWanModel(WanModel):
 
         self.patch_embedding_pose = operations.Conv3d(in_dim, dim, kernel_size=patch_size, stride=patch_size, device=device, dtype=torch.float32)
 
-    def forward_orig(self, x, t, context, clip_fea=None, freqs=None, transformer_options={}, pose_latents=None, reference_latent=None, **kwargs):
+    def forward_orig(self, x, t, context, clip_fea=None, freqs=None, transformer_options={}, pose_latents=None, reference_latent=None, ref_mask_latents=None, sam_latents=None, **kwargs):
 
         if reference_latent is not None:
             x = torch.cat((reference_latent, x), dim=2)
 
         # embeddings
         x = self.patch_embedding(x.float()).to(x.dtype)
+        if ref_mask_latents is not None:  # SCAIL-2 additive mask stream (one identity mask frame per reference, then video)
+            x = x + self.patch_embedding_mask(ref_mask_latents.float()).to(x.dtype)
         grid_sizes = x.shape[2:]
         transformer_options["grid_sizes"] = grid_sizes
         x = x.flatten(2).transpose(1, 2)
@@ -1645,6 +1674,8 @@ class SCAILWanModel(WanModel):
         scail_pose_seq_len = 0
         if pose_latents is not None:
             scail_x = self.patch_embedding_pose(pose_latents.float()).to(x.dtype)
+            if sam_latents is not None:  # SCAIL-2 additive mask stream
+                scail_x = scail_x + self.patch_embedding_mask(sam_latents.float()).to(x.dtype)
             scail_x = scail_x.flatten(2).transpose(1, 2)
             scail_pose_seq_len = scail_x.shape[1]
             x = torch.cat([x, scail_x], dim=1)
@@ -1695,16 +1726,44 @@ class SCAILWanModel(WanModel):
 
         return x
 
-    def rope_encode(self, t, h, w, t_start=0, steps_t=None, steps_h=None, steps_w=None, device=None, dtype=None, pose_latents=None, reference_latent=None, transformer_options={}):
+    # ref_mask_flag is a scalar bool (CONDConstant, SCAIL-2 only). False => replacement mode,
+    # which places ref/pose via H/W rope shifts instead of the animation-mode temporal offset.
+    # reference_latent may stack several frames: the last is the primary reference adjacent to the video, the earlier frames are additional references.
+    def rope_encode(self, t, h, w, t_start=0, steps_t=None, steps_h=None, steps_w=None, device=None, dtype=None, pose_latents=None, reference_latent=None, ref_mask_flag=None, transformer_options={}):
+        ref_t_patches = 0
+        if reference_latent is not None:
+            ref_t_patches = (reference_latent.shape[2] + (self.patch_size[0] // 2)) // self.patch_size[0]
+
+        if ref_mask_flag is not None and not bool(ref_mask_flag):
+            REF_ROPE_H = 120.0
+            POSE_ROPE_W = 120.0
+
+            main_t_patches = t - ref_t_patches
+            video_t_start = max(ref_t_patches - 1, 0)
+
+            parts = []
+            if ref_t_patches > 0:
+                ref_tf = {"rope_options": {"shift_y": REF_ROPE_H, "shift_x": 0.0, "scale_y": 1.0, "scale_x": 1.0}}
+                parts.append(super().rope_encode(ref_t_patches, h, w, t_start=0, device=device, dtype=dtype, transformer_options=ref_tf))
+            if main_t_patches > 0:
+                parts.append(super().rope_encode(main_t_patches, h, w, t_start=video_t_start, device=device, dtype=dtype, transformer_options=transformer_options))
+
+            if pose_latents is not None:
+                F_pose, H_pose, W_pose = pose_latents.shape[-3], pose_latents.shape[-2], pose_latents.shape[-1]
+                h_scale = h / H_pose
+                w_scale = w / W_pose
+                h_shift = (h_scale - 1) / 2
+                w_shift = (w_scale - 1) / 2
+                pose_tf = {"rope_options": {"shift_y": h_shift, "shift_x": POSE_ROPE_W + w_shift, "scale_y": h_scale, "scale_x": w_scale}}
+                parts.append(super().rope_encode(F_pose, H_pose, W_pose, t_start=video_t_start, device=device, dtype=dtype, transformer_options=pose_tf))
+
+            return torch.cat(parts, dim=1)
+
         main_freqs = super().rope_encode(t, h, w, t_start=t_start, steps_t=steps_t, steps_h=steps_h, steps_w=steps_w, device=device, dtype=dtype, transformer_options=transformer_options)
 
         if pose_latents is None:
             return main_freqs
 
-        ref_t_patches = 0
-        if reference_latent is not None:
-            ref_t_patches = (reference_latent.shape[2] + (self.patch_size[0] // 2)) // self.patch_size[0]
-
         F_pose, H_pose, W_pose = pose_latents.shape[-3], pose_latents.shape[-2], pose_latents.shape[-1]
 
         # if pose is at half resolution, scale_y/scale_x=2 stretches the position range to cover the same RoPE extent as the main frames
@@ -1719,12 +1778,16 @@ class SCAILWanModel(WanModel):
 
         return torch.cat([main_freqs, pose_freqs], dim=1)
 
-    def _forward(self, x, timestep, context, clip_fea=None, time_dim_concat=None, transformer_options={}, pose_latents=None, **kwargs):
+    def _forward(self, x, timestep, context, clip_fea=None, time_dim_concat=None, transformer_options={}, pose_latents=None, ref_mask_latents=None, sam_latents=None, **kwargs):
         bs, c, t, h, w = x.shape
         x = comfy.ldm.common_dit.pad_to_patch_size(x, self.patch_size)
 
         if pose_latents is not None:
             pose_latents = comfy.ldm.common_dit.pad_to_patch_size(pose_latents, self.patch_size)
+        if ref_mask_latents is not None:  # SCAIL-2
+            ref_mask_latents = comfy.ldm.common_dit.pad_to_patch_size(ref_mask_latents, self.patch_size)
+        if sam_latents is not None:  # SCAIL-2
+            sam_latents = comfy.ldm.common_dit.pad_to_patch_size(sam_latents, self.patch_size)
 
         t_len = t
         if time_dim_concat is not None:
@@ -1737,5 +1800,15 @@ class SCAILWanModel(WanModel):
             reference_latent = comfy.ldm.common_dit.pad_to_patch_size(kwargs.pop("reference_latent"), self.patch_size)
             t_len += reference_latent.shape[2]
 
-        freqs = self.rope_encode(t_len, h, w, device=x.device, dtype=x.dtype, transformer_options=transformer_options, pose_latents=pose_latents, reference_latent=reference_latent)
-        return self.forward_orig(x, timestep, context, clip_fea=clip_fea, freqs=freqs, transformer_options=transformer_options, pose_latents=pose_latents, reference_latent=reference_latent, **kwargs)[:, :, :t, :h, :w]
+        ref_mask_flag = kwargs.pop("ref_mask_flag", None)  # SCAIL-2
+
+        freqs = self.rope_encode(t_len, h, w, device=x.device, dtype=x.dtype, transformer_options=transformer_options, pose_latents=pose_latents, reference_latent=reference_latent, ref_mask_flag=ref_mask_flag)
+        return self.forward_orig(x, timestep, context, clip_fea=clip_fea, freqs=freqs, transformer_options=transformer_options, pose_latents=pose_latents, reference_latent=reference_latent, ref_mask_latents=ref_mask_latents, sam_latents=sam_latents, **kwargs)[:, :, :t, :h, :w]
+
+
+class SCAIL2WanModel(SCAILWanModel):
+    """SCAIL-2: SCAIL-Preview + an additive binary multi-identity mask stream."""
+
+    def __init__(self, model_type="scail2", patch_size=(1, 2, 2), in_dim=20, mask_in_dim=28, dim=5120, operations=None, device=None, dtype=None, **kwargs):
+        super().__init__(model_type=model_type, patch_size=patch_size, in_dim=in_dim, dim=dim, operations=operations, device=device, dtype=dtype, **kwargs)
+        self.patch_embedding_mask = operations.Conv3d(mask_in_dim, dim, kernel_size=patch_size, stride=patch_size, device=device, dtype=torch.float32)

comfy/lora.py

@@ -357,6 +357,12 @@ def model_lora_keys_unet(model, key_map={}):
                 key_lora = k[len("diffusion_model."):-len(".weight")]
                 key_map["transformer.{}".format(key_lora)] = k
 
+    if isinstance(model, (comfy.model_base.LTXV, comfy.model_base.LTXAV)):
+        for k in sdk:
+            if k.startswith("diffusion_model.") and k.endswith(".weight"):
+                key_lora = k[len("diffusion_model."):-len(".weight")]
+                key_map["{}".format(key_lora)] = k
+
     return key_map
 
 

comfy/memory_management.py

@@ -4,6 +4,7 @@ import dataclasses
 import torch
 from typing import NamedTuple
 
+import comfy_aimdo.host_buffer
 from comfy.quant_ops import QuantizedTensor
 
 
@@ -17,21 +18,18 @@ class TensorFileSlice(NamedTuple):
 def read_tensor_file_slice_into(tensor, destination, stream=None, destination2=None):
 
     if isinstance(tensor, QuantizedTensor):
-        if not isinstance(destination, QuantizedTensor):
-            return False
-        if tensor._layout_cls != destination._layout_cls:
-            return False
-
-        if not read_tensor_file_slice_into(tensor._qdata, destination._qdata, stream=stream,
+        if not read_tensor_file_slice_into(tensor._qdata,
+                                           destination._qdata if destination is not None else None, stream=stream,
                                            destination2=(destination2._qdata if destination2 is not None else None)):
             return False
 
-        dst_orig_dtype = destination._params.orig_dtype
-        destination._params.copy_from(tensor._params, non_blocking=False)
-        destination._params = dataclasses.replace(destination._params, orig_dtype=dst_orig_dtype)
+        if destination is not None:
+            dst_orig_dtype = destination._params.orig_dtype
+            destination._params.copy_from(tensor._params, non_blocking=False)
+            destination._params = dataclasses.replace(destination._params, orig_dtype=dst_orig_dtype)
         if destination2 is not None:
             dst_orig_dtype = destination2._params.orig_dtype
-            destination2._params.copy_from(destination._params, non_blocking=True)
+            destination2._params.copy_from(destination._params if destination is not None else tensor._params, non_blocking=True)
             destination2._params = dataclasses.replace(destination2._params, orig_dtype=dst_orig_dtype)
         return True
 
@@ -39,10 +37,15 @@ def read_tensor_file_slice_into(tensor, destination, stream=None, destination2=N
     if info is None:
         return False
 
+    if destination is not None and destination.device.type != "cpu" and destination2 is None:
+        destination2 = destination
+        destination = None
+
     file_obj = info.file_ref
-    if (destination.device.type != "cpu"
-            or file_obj is None
-            or destination.numel() * destination.element_size() < info.size
+    if (file_obj is None
+            or (destination is None and destination2 is None)
+            or (destination is not None and (destination.device.type != "cpu" or destination.numel() * destination.element_size() < info.size))
+            or (destination2 is not None and (destination2.device.type == "cpu" or destination2.numel() * destination2.element_size() < info.size))
             or tensor.numel() * tensor.element_size() != info.size
             or tensor.storage_offset() != 0
             or not tensor.is_contiguous()):
@@ -51,6 +54,14 @@ def read_tensor_file_slice_into(tensor, destination, stream=None, destination2=N
     if info.size == 0:
         return True
 
+    if destination is None:
+        stream_ptr = getattr(stream, "cuda_stream", 0) if stream is not None else 0
+        comfy_aimdo.host_buffer.read_file_to_device(file_obj, info.offset, info.size,
+                                                    stream_ptr, destination2.data_ptr(),
+                                                    destination2.device.index,
+                                                    mark_cold=False)
+        return True
+
     hostbuf = getattr(destination.untyped_storage(), "_comfy_hostbuf", None)
     if hostbuf is not None:
         stream_ptr = getattr(stream, "cuda_stream", 0) if stream is not None else 0
@@ -63,6 +74,9 @@ def read_tensor_file_slice_into(tensor, destination, stream=None, destination2=N
                                     device=None if destination2 is None else destination2.device.index)
         return True
 
+    if not hasattr(file_obj, "seek") or not hasattr(file_obj, "readinto"):
+        return False
+
     buf_type = ctypes.c_ubyte * info.size
     view = memoryview(buf_type.from_address(destination.data_ptr()))
 

comfy/model_base.py

@@ -21,6 +21,7 @@ import comfy.ldm.hunyuan3dv2_1.hunyuandit
 import torch
 import logging
 import comfy.ldm.lightricks.av_model
+import comfy.ldm.lightricks.symmetric_patchifier
 import comfy.context_windows
 from comfy.ldm.modules.diffusionmodules.openaimodel import UNetModel, Timestep
 from comfy.ldm.cascade.stage_c import StageC
@@ -46,6 +47,7 @@ import comfy.ldm.wan.model_animate
 import comfy.ldm.wan.ar_model
 import comfy.ldm.wan.model_wandancer
 import comfy.ldm.hunyuan3d.model
+import comfy.ldm.triposplat.model
 import comfy.ldm.hidream.model
 import comfy.ldm.chroma.model
 import comfy.ldm.chroma_radiance.model
@@ -53,7 +55,9 @@ import comfy.ldm.pixeldit.model
 import comfy.ldm.pixeldit.pid
 import comfy.ldm.ace.model
 import comfy.ldm.omnigen.omnigen2
+import comfy.ldm.boogu.model
 import comfy.ldm.qwen_image.model
+import comfy.ldm.ideogram4.model
 import comfy.ldm.kandinsky5.model
 import comfy.ldm.anima.model
 import comfy.ldm.ace.ace_step15
@@ -63,6 +67,7 @@ import comfy.ldm.ernie.model
 import comfy.ldm.sam3.detector
 import comfy.ldm.hidream_o1.model
 from comfy.ldm.hidream_o1.conditioning import build_extra_conds
+import comfy.ldm.depth_anything_3.model
 
 import comfy.model_management
 import comfy.patcher_extension
@@ -1200,6 +1205,127 @@ class LTXAV(BaseModel):
     def scale_latent_inpaint(self, sigma, noise, latent_image, **kwargs):
         return latent_image
 
+    def map_context_window_to_modalities(self, primary_indices, latent_shapes, dim):
+        result = [primary_indices]
+        if len(latent_shapes) < 2:
+            return result
+
+        video_total = latent_shapes[0][dim]
+
+        for i in range(1, len(latent_shapes)):
+            mod_total = latent_shapes[i][dim]
+            # Map each primary index to its proportional range of modality indices and
+            # concatenate in order. Preserves wrapped/strided geometry so the modality
+            # attends to the same temporal regions as the primary window.
+            mod_indices = []
+            seen = set()
+            for v_idx in primary_indices:
+                a_start = min(int(round(v_idx * mod_total / video_total)), mod_total - 1)
+                a_end = min(int(round((v_idx + 1) * mod_total / video_total)), mod_total)
+                if a_end <= a_start:
+                    a_end = a_start + 1
+                for a in range(a_start, a_end):
+                    if a not in seen:
+                        seen.add(a)
+                        mod_indices.append(a)
+            result.append(mod_indices)
+
+        return result
+
+    @staticmethod
+    def _get_guide_entries(conds):
+        for cond_list in conds:
+            if cond_list is None:
+                continue
+            for cond_dict in cond_list:
+                model_conds = cond_dict.get('model_conds', {})
+                entries = model_conds.get('guide_attention_entries')
+                if entries is not None and hasattr(entries, 'cond') and entries.cond:
+                    return entries.cond
+        return None
+
+    def resize_cond_for_context_window(self, cond_key, cond_value, window, x_in, device, retain_index_list=[]):
+        # Audio denoise mask — slice using audio modality window
+        if cond_key == "audio_denoise_mask" and hasattr(window, 'modality_windows') and window.modality_windows:
+            audio_window = window.modality_windows.get(1)
+            if audio_window is not None and hasattr(cond_value, "cond") and isinstance(cond_value.cond, torch.Tensor):
+                sliced = audio_window.get_tensor(cond_value.cond, device, dim=2)
+                return cond_value._copy_with(sliced)
+
+        # Video denoise mask — split into video + guide portions, slice each
+        if cond_key == "denoise_mask" and hasattr(cond_value, "cond") and isinstance(cond_value.cond, torch.Tensor):
+            cond_tensor = cond_value.cond
+            guide_count = cond_tensor.size(window.dim) - x_in.size(window.dim)
+            if guide_count > 0:
+                T_video = x_in.size(window.dim)
+                video_mask = cond_tensor.narrow(window.dim, 0, T_video)
+                guide_mask = cond_tensor.narrow(window.dim, T_video, guide_count)
+                sliced_video = window.get_tensor(video_mask, device, retain_index_list=retain_index_list)
+                suffix_indices = window.guide_frames_indices
+                if suffix_indices:
+                    idx = tuple([slice(None)] * window.dim + [suffix_indices])
+                    sliced_guide = guide_mask[idx].to(device)
+                    return cond_value._copy_with(torch.cat([sliced_video, sliced_guide], dim=window.dim))
+                else:
+                    return cond_value._copy_with(sliced_video)
+
+        # Keyframe indices — regenerate pixel coords for window, select guide positions
+        if cond_key == "keyframe_idxs":
+            kf_local_pos = window.guide_kf_local_positions
+            if not kf_local_pos:
+                return cond_value._copy_with(cond_value.cond[:, :, :0, :])  # empty
+            H, W = x_in.shape[3], x_in.shape[4]
+            window_len = len(window.index_list)
+            # account for causal_window_fix anchor in coord space size
+            anchor_idx = getattr(window, 'causal_anchor_index', None)
+            if anchor_idx is not None and anchor_idx >= 0:
+                window_len += 1
+            patchifier = self.diffusion_model.patchifier
+            latent_coords = patchifier.get_latent_coords(window_len, H, W, 1, cond_value.cond.device)
+            scale_factors = self.diffusion_model.vae_scale_factors
+            pixel_coords = comfy.ldm.lightricks.symmetric_patchifier.latent_to_pixel_coords(
+                latent_coords,
+                scale_factors,
+                causal_fix=self.diffusion_model.causal_temporal_positioning)
+            tokens = []
+            for pos in kf_local_pos:
+                tokens.extend(range(pos * H * W, (pos + 1) * H * W))
+            pixel_coords = pixel_coords[:, :, tokens, :]
+
+            # Adjust spatial end positions for dilated (downscaled) guides.
+            # Each guide entry may have a different downscale factor; expand the
+            # per-entry factor to cover all tokens belonging to that entry.
+            downscale_factors = window.guide_downscale_factors
+            overlap_info = window.guide_overlap_info
+            if downscale_factors:
+                per_token_factor = []
+                for (entry_idx, overlap_count), dsf in zip(overlap_info, downscale_factors):
+                    per_token_factor.extend([dsf] * (overlap_count * H * W))
+                factor_tensor = torch.tensor(per_token_factor, device=pixel_coords.device, dtype=pixel_coords.dtype)
+                spatial_end_offset = (factor_tensor.unsqueeze(0).unsqueeze(0).unsqueeze(-1) - 1) * torch.tensor(
+                    scale_factors[1:], device=pixel_coords.device, dtype=pixel_coords.dtype,
+                ).view(1, -1, 1, 1)
+                pixel_coords[:, 1:, :, 1:] += spatial_end_offset
+
+            B = cond_value.cond.shape[0]
+            if B > 1:
+                pixel_coords = pixel_coords.expand(B, -1, -1, -1)
+            return cond_value._copy_with(pixel_coords)
+
+        # Guide attention entries — adjust per-guide counts based on window overlap
+        if cond_key == "guide_attention_entries":
+            overlap_info = window.guide_overlap_info
+            H, W = x_in.shape[3], x_in.shape[4]
+            new_entries = []
+            for entry_idx, overlap_count in overlap_info:
+                e = cond_value.cond[entry_idx]
+                new_entries.append({**e,
+                    "pre_filter_count": overlap_count * H * W,
+                    "latent_shape": [overlap_count, H, W]})
+            return cond_value._copy_with(new_entries)
+
+        return None
+
 class HunyuanVideo(BaseModel):
     def __init__(self, model_config, model_type=ModelType.FLOW, device=None):
         super().__init__(model_config, model_type, device=device, unet_model=comfy.ldm.hunyuan_video.model.HunyuanVideo)
@@ -1516,8 +1642,26 @@ class WAN21(BaseModel):
         if reference_latents is not None:
             out['reference_latent'] = comfy.conds.CONDRegular(self.process_latent_in(reference_latents[-1])[:, :, 0])
 
+        # In-context reference conditioning (Bernini)
+        context_latents = kwargs.get("context_latents", None)
+        if context_latents is not None:
+            out['context_latents'] = comfy.conds.CONDList([self.process_latent_in(l) for l in context_latents])
+
         return out
 
+    def resize_cond_for_context_window(self, cond_key, cond_value, window, x_in, device, retain_index_list=[]):
+        # In-context cond slicing (Bernini)
+        if cond_key == "context_latents" and isinstance(getattr(cond_value, "cond", None), list):
+            dim = window.dim
+            out = []
+            for lat in cond_value.cond:
+                if lat.ndim > dim and lat.shape[dim] > 1 and lat.shape[dim] == x_in.shape[dim]:
+                    out.append(window.get_tensor(lat, device, dim=dim, retain_index_list=retain_index_list))
+                else:
+                    out.append(lat.to(device))
+            return cond_value._copy_with(out)
+        return super().resize_cond_for_context_window(cond_key, cond_value, window, x_in, device, retain_index_list=retain_index_list)
+
 
 class WAN21_CausalAR(WAN21):
     def __init__(self, model_config, model_type=ModelType.FLOW, device=None):
@@ -1726,10 +1870,14 @@ class WAN21_SCAIL(WAN21):
 
         reference_latents = kwargs.get("reference_latents", None)
         if reference_latents is not None:
-            ref_latent = self.process_latent_in(reference_latents[-1])
-            ref_mask = torch.ones_like(ref_latent[:, :4])
-            ref_latent = torch.cat([ref_latent, ref_mask], dim=1)
-            out['reference_latent'] = comfy.conds.CONDRegular(ref_latent)
+            # SCAIL-2 multi-reference: reference_latents[0] is the primary ref, [1:] are additional
+            # references. Stack as [additional..., primary] so the primary stays adjacent to the video.
+            ordered = list(reference_latents[1:]) + list(reference_latents[:1])
+            stacked = []
+            for lat in ordered:
+                lat = self.process_latent_in(lat)
+                stacked.append(torch.cat([lat, torch.ones_like(lat[:, :4])], dim=1))
+            out['reference_latent'] = comfy.conds.CONDRegular(torch.cat(stacked, dim=2))
 
         pose_latents = kwargs.get("pose_video_latent", None)
         if pose_latents is not None:
@@ -1752,6 +1900,99 @@ class WAN21_SCAIL(WAN21):
 
         return out
 
+class WAN21_SCAIL2(WAN21_SCAIL):
+    """SCAIL-2: SCAIL-Preview + an additive binary multi-identity mask stream."""
+
+    def __init__(self, model_config, model_type=ModelType.FLOW, image_to_video=False, device=None):
+        super(WAN21, self).__init__(model_config, model_type, device=device, unet_model=comfy.ldm.wan.model.SCAIL2WanModel)
+        self.memory_usage_factor_conds = ("reference_latent", "pose_latents", "ref_mask_latents", "sam_latents")
+        self.memory_usage_shape_process = {
+            "pose_latents": lambda shape: [shape[0], shape[1], 1.5, shape[-2], shape[-1]],
+            "sam_latents":  lambda shape: [shape[0], shape[1], 1.5, shape[-2], shape[-1]],
+        }
+        self.image_to_video = image_to_video
+
+    def extra_conds(self, **kwargs):
+        out = super().extra_conds(**kwargs)
+
+        driving_mask_28ch = kwargs.get("driving_mask_28ch", None)
+        if driving_mask_28ch is not None:
+            out['sam_latents'] = comfy.conds.CONDRegular(driving_mask_28ch.movedim(1, 2).contiguous())
+
+        # ref_mask_28ch holds one identity mask per stacked reference frame (additional refs first, then the primary ref), followed by zeros over the video frames.
+        ref_mask_28ch = kwargs.get("ref_mask_28ch", None)
+        if ref_mask_28ch is not None:
+            out['ref_mask_latents'] = comfy.conds.CONDRegular(ref_mask_28ch.movedim(1, 2).contiguous())
+
+        ref_mask_flag = kwargs.get("ref_mask_flag", None)
+        if ref_mask_flag is not None:
+            out['ref_mask_flag'] = comfy.conds.CONDConstant(ref_mask_flag)
+
+        return out
+
+    def extra_conds_shapes(self, **kwargs):
+        out = super().extra_conds_shapes(**kwargs)
+        driving_mask_28ch = kwargs.get("driving_mask_28ch", None)
+        if driving_mask_28ch is not None:
+            s = driving_mask_28ch.shape
+            out['sam_latents'] = [s[0], 28, s[1], s[3], s[4]]
+        ref_mask_28ch = kwargs.get("ref_mask_28ch", None)
+        if ref_mask_28ch is not None:
+            s = ref_mask_28ch.shape
+            out['ref_mask_latents'] = [s[0], 28, s[1], s[3], s[4]]
+        return out
+
+    def resize_cond_for_context_window(self, cond_key, cond_value, window, x_in, device, retain_index_list=[]):
+        if cond_key in ("sam_latents", "pose_latents"):
+            # Return sliced view omitting retain_index_list
+            return comfy.context_windows.slice_cond(cond_value, window, x_in, device, temporal_dim=2, temporal_offset=0)
+        if cond_key == "ref_mask_latents" and hasattr(cond_value, "cond") and isinstance(cond_value.cond, torch.Tensor):
+            # The ref mask is N leading ref frames padded with frames of zeros, so just grab the first frames for all windows
+            full_ref_mask = cond_value.cond
+            video_frame_count = x_in.shape[2]
+            ref_frame_count = full_ref_mask.shape[2] - video_frame_count
+            if ref_frame_count < 1:
+                return None
+            window_length = len(window.index_list)
+
+            # Account for the causal anchor frame if it exists
+            anchor_index = getattr(window, "causal_anchor_index", None)
+            if anchor_index is not None and anchor_index >= 0:
+                window_length += 1
+
+            window_ref_mask = full_ref_mask[:, :, :window_length + ref_frame_count].to(device)
+            return cond_value._copy_with(window_ref_mask)
+
+        return super().resize_cond_for_context_window(cond_key, cond_value, window, x_in, device, retain_index_list=retain_index_list)
+
+    def concat_cond(self, **kwargs):
+        # The 4 extra channels are the history_mask (1 at clean-anchor frames).
+        noise = kwargs.get("noise", None)
+        extra_channels = self.diffusion_model.patch_embedding.weight.shape[1] - noise.shape[1]
+        if extra_channels != 4:
+            return super().concat_cond(**kwargs)
+
+        mask = kwargs.get("concat_mask", kwargs.get("denoise_mask", None))
+        if mask is None:
+            return torch.zeros_like(noise)[:, :4]
+
+        device = kwargs["device"]
+        if mask.shape[1] != 4:
+            mask = torch.mean(mask, dim=1, keepdim=True)
+        mask = 1.0 - mask
+        mask = utils.common_upscale(mask.to(device), noise.shape[-1], noise.shape[-2], "bilinear", "center")
+        if mask.shape[-3] < noise.shape[-3]:
+            mask = torch.nn.functional.pad(mask, (0, 0, 0, 0, 0, noise.shape[-3] - mask.shape[-3]), mode='constant', value=0)
+        if mask.shape[1] == 1:
+            mask = mask.repeat(1, 4, 1, 1, 1)
+        mask = utils.resize_to_batch_size(mask, noise.shape[0])
+        return mask
+
+    def scale_latent_inpaint(self, sigma, noise, latent_image, **kwargs):
+        # Hold anchor constant across all sigmas instead of base sigma*noise + (1-sigma)*latent_image.
+        return latent_image
+
+
 class WAN22_WanDancer(WAN21):
     def __init__(self, model_config, model_type=ModelType.FLOW, image_to_video=True, device=None):
         super(WAN21, self).__init__(model_config, model_type, device=device, unet_model=comfy.ldm.wan.model_wandancer.WanDancerModel)
@@ -1806,6 +2047,24 @@ class Hunyuan3Dv2_1(BaseModel):
             out['guidance'] = comfy.conds.CONDRegular(torch.FloatTensor([guidance]))
         return out
 
+class TripoSplat(BaseModel):
+    def __init__(self, model_config, model_type=ModelType.FLOW, device=None):
+        super().__init__(model_config, model_type, device=device, unet_model=comfy.ldm.triposplat.model.LatentSeqMMFlowModel)
+
+    def extra_conds(self, **kwargs):
+        out = super().extra_conds(**kwargs)
+        cross_attn = kwargs.get("cross_attn", None) # DINOv3 token sequence -> cross-attention context.
+        if cross_attn is not None:
+            out['c_crossattn'] = comfy.conds.CONDRegular(cross_attn)
+        ref_latents = kwargs.get("reference_latents", None) # Flux2 VAE image latent -> additive second conditioning.
+        if ref_latents is not None:
+            out['ref_latents'] = comfy.conds.CONDList(list(ref_latents))
+        latent_shapes = kwargs.get("latent_shapes", None) # {latent, camera} nested latent
+        if latent_shapes is not None:
+            out['latent_shapes'] = comfy.conds.CONDConstant(latent_shapes)
+        return out
+
+
 class HiDream(BaseModel):
     def __init__(self, model_config, model_type=ModelType.FLOW, device=None):
         super().__init__(model_config, model_type, device=device, unet_model=comfy.ldm.hidream.model.HiDreamImageTransformer2DModel)
@@ -1967,6 +2226,11 @@ class Omnigen2(BaseModel):
             out['ref_latents'] = list([1, 16, sum(map(lambda a: math.prod(a.size()), ref_latents)) // 16])
         return out
 
+class Boogu(Omnigen2):
+    def __init__(self, model_config, model_type=ModelType.FLOW, device=None):
+        super(Omnigen2, self).__init__(model_config, model_type, device=device, unet_model=comfy.ldm.boogu.model.BooguTransformer2DModel)
+        self.memory_usage_factor_conds = ("ref_latents",)
+
 class QwenImage(BaseModel):
     def __init__(self, model_config, model_type=ModelType.FLUX, device=None):
         super().__init__(model_config, model_type, device=device, unet_model=comfy.ldm.qwen_image.model.QwenImageTransformer2DModel)
@@ -1999,6 +2263,21 @@ class QwenImage(BaseModel):
             out['ref_latents'] = list([1, 16, sum(map(lambda a: math.prod(a.size()), ref_latents)) // 16])
         return out
 
+class Ideogram4(BaseModel):
+    def __init__(self, model_config, model_type=ModelType.FLOW, device=None):
+        super().__init__(model_config, model_type, device=device, unet_model=comfy.ldm.ideogram4.model.Ideogram4Transformer2DModel)
+
+    def extra_conds(self, **kwargs):
+        out = super().extra_conds(**kwargs)
+        attention_mask = kwargs.get("attention_mask", None)
+        if attention_mask is not None:
+            if torch.numel(attention_mask) != attention_mask.sum():
+                out['attention_mask'] = comfy.conds.CONDRegular(attention_mask)
+        cross_attn = kwargs.get("cross_attn", None)
+        if cross_attn is not None:
+            out['c_crossattn'] = comfy.conds.CONDRegular(cross_attn)
+        return out
+
 class HunyuanImage21(BaseModel):
     def __init__(self, model_config, model_type=ModelType.FLOW, device=None):
         super().__init__(model_config, model_type, device=device, unet_model=comfy.ldm.hunyuan_video.model.HunyuanVideo)
@@ -2192,6 +2471,12 @@ class RT_DETR_v4(BaseModel):
     def __init__(self, model_config, model_type=ModelType.FLOW, device=None):
         super().__init__(model_config, model_type, device=device, unet_model=comfy.ldm.rt_detr.rtdetr_v4.RTv4)
 
+
+class DepthAnything3(BaseModel):
+    def __init__(self, model_config, model_type=ModelType.FLOW, device=None):
+        super().__init__(model_config, model_type, device=device,
+                         unet_model=comfy.ldm.depth_anything_3.model.DepthAnything3Net)
+
 class ErnieImage(BaseModel):
     def __init__(self, model_config, model_type=ModelType.FLOW, device=None):
         super().__init__(model_config, model_type, device=device, unet_model=comfy.ldm.ernie.model.ErnieImageModel)

comfy/model_detection.py

@@ -313,6 +313,10 @@ def detect_unet_config(state_dict, key_prefix, metadata=None):
                     dit_config["use_x0"] = True
                 else:
                     dit_config["use_x0"] = False
+                if "{}__sequential__".format(key_prefix) in state_dict_keys: # sequential txt_ids
+                    dit_config["use_sequential_txt_ids"] = True
+                else:
+                    dit_config["use_sequential_txt_ids"] = False
         else:
             dit_config["guidance_embed"] = "{}guidance_in.in_layer.weight".format(key_prefix) in state_dict_keys
             dit_config["yak_mlp"] = '{}double_blocks.0.img_mlp.gate_proj.weight'.format(key_prefix) in state_dict_keys
@@ -626,6 +630,8 @@ def detect_unet_config(state_dict, key_prefix, metadata=None):
             dit_config["model_type"] = "humo"
         elif '{}face_adapter.fuser_blocks.0.k_norm.weight'.format(key_prefix) in state_dict_keys:
             dit_config["model_type"] = "animate"
+        elif '{}patch_embedding_mask.weight'.format(key_prefix) in state_dict_keys:
+            dit_config["model_type"] = "scail2"
         elif '{}patch_embedding_pose.weight'.format(key_prefix) in state_dict_keys:
             dit_config["model_type"] = "scail"
         elif '{}patch_embedding_global.weight'.format(key_prefix) in state_dict_keys:
@@ -676,6 +682,9 @@ def detect_unet_config(state_dict, key_prefix, metadata=None):
         dit_config["guidance_cond_proj_dim"] = None#f"{key_prefix}t_embedder.cond_proj.weight" in state_dict_keys
         return dit_config
 
+    if '{}cam_out_layer.weight'.format(key_prefix) in state_dict_keys and '{}repo_layers.0.final_map.weight'.format(key_prefix) in state_dict_keys:  # TripoSplat
+        return {"image_model": "triposplat"}
+
     if '{}t_embedder1.mlp.0.weight'.format(key_prefix) in state_dict_keys and '{}x_embedder.proj1.weight'.format(key_prefix) in state_dict_keys:  # HiDream-O1
         return {"image_model": "hidream_o1"}
 
@@ -752,6 +761,16 @@ def detect_unet_config(state_dict, key_prefix, metadata=None):
 
         return dit_config
 
+    if '{}double_stream_layers.0.img_instruct_attn.processor.img_to_q.weight'.format(key_prefix) in state_dict_keys:  # Boogu-Image (OmniGen2 derivative + dual-stream stage)
+        dit_config = {}
+        dit_config["image_model"] = "boogu"
+        dit_config["hidden_size"] = state_dict['{}x_embedder.weight'.format(key_prefix)].shape[0]
+        dit_config["num_layers"] = count_blocks(state_dict_keys, '{}single_stream_layers.'.format(key_prefix) + '{}.')
+        dit_config["num_double_stream_layers"] = count_blocks(state_dict_keys, '{}double_stream_layers.'.format(key_prefix) + '{}.')
+        dit_config["num_refiner_layers"] = count_blocks(state_dict_keys, '{}noise_refiner.'.format(key_prefix) + '{}.')
+        dit_config["instruction_feat_dim"] = state_dict['{}time_caption_embed.caption_embedder.0.weight'.format(key_prefix)].shape[0]
+        return dit_config
+
     if '{}time_caption_embed.timestep_embedder.linear_1.bias'.format(key_prefix) in state_dict_keys:  # Omnigen2
         dit_config = {}
         dit_config["image_model"] = "omnigen2"
@@ -808,6 +827,13 @@ def detect_unet_config(state_dict, key_prefix, metadata=None):
             dit_config["default_ref_method"] = "negative_index"
         return dit_config
 
+    if '{}embed_image_indicator.weight'.format(key_prefix) in state_dict_keys:  # Ideogram 4
+        dit_config = {}
+        dit_config["image_model"] = "ideogram4"
+        dit_config["in_channels"] = state_dict['{}input_proj.weight'.format(key_prefix)].shape[1]
+        dit_config["num_layers"] = count_blocks(state_dict_keys, '{}layers.'.format(key_prefix) + '{}.')
+        return dit_config
+
     if '{}visual_transformer_blocks.0.cross_attention.key_norm.weight'.format(key_prefix) in state_dict_keys: # Kandinsky 5
         dit_config = {}
         model_dim = state_dict['{}visual_embeddings.in_layer.bias'.format(key_prefix)].shape[0]
@@ -846,6 +872,95 @@ def detect_unet_config(state_dict, key_prefix, metadata=None):
         dit_config["enc_h"] = state_dict['{}encoder.pan_blocks.1.cv4.conv.weight'.format(key_prefix)].shape[0]
         return dit_config
 
+    # Depth Anything 3 (repackaged to ComfyUI's native Dinov2Model layout via scripts/convert_da3.py)
+    if '{}backbone.embeddings.patch_embeddings.projection.weight'.format(key_prefix) in state_dict_keys:
+        dit_config = {}
+        dit_config["image_model"] = "DepthAnything3"
+
+        patch_w = state_dict['{}backbone.embeddings.patch_embeddings.projection.weight'.format(key_prefix)]
+        embed_dim = patch_w.shape[0]
+        depth = count_blocks(state_dict_keys, '{}backbone.encoder.layer.'.format(key_prefix) + '{}.')
+
+        # Backbone preset is determined by embed_dim (matches vits/vitb/vitl/vitg).
+        backbone_name = {384: "vits", 768: "vitb", 1024: "vitl", 1536: "vitg"}.get(embed_dim)
+        if backbone_name is None:
+            return None
+        dit_config["backbone_name"] = backbone_name
+
+        # Detect DA3 extensions on top of vanilla DINOv2.
+        has_camera_token = '{}backbone.embeddings.camera_token'.format(key_prefix) in state_dict_keys
+        # qk-norm shows up as `attention.q_norm.weight` on enabled blocks.
+        qknorm_indices = [
+            i for i in range(depth)
+            if '{}backbone.encoder.layer.{}.attention.q_norm.weight'.format(key_prefix, i) in state_dict_keys
+        ]
+        qknorm_start = qknorm_indices[0] if qknorm_indices else -1
+
+        # The DA3 main-series configs always set alt_start == qknorm_start == rope_start.
+        # cat_token=True is implied by the presence of camera_token.
+        if has_camera_token:
+            dit_config["alt_start"] = qknorm_start
+            dit_config["rope_start"] = qknorm_start
+            dit_config["qknorm_start"] = qknorm_start
+            dit_config["cat_token"] = True
+        else:
+            dit_config["alt_start"] = -1
+            dit_config["rope_start"] = -1
+            dit_config["qknorm_start"] = -1
+            dit_config["cat_token"] = False
+
+        # Detect head type and config.
+        has_aux = '{}head.scratch.refinenet1_aux.out_conv.weight'.format(key_prefix) in state_dict_keys
+        dit_config["head_dim_in"] = state_dict['{}head.projects.0.weight'.format(key_prefix)].shape[1]
+        dit_config["head_features"] = state_dict['{}head.scratch.refinenet1.out_conv.weight'.format(key_prefix)].shape[0]
+        dit_config["head_out_channels"] = [
+            state_dict['{}head.projects.{}.weight'.format(key_prefix, i)].shape[0]
+            for i in range(4)
+        ]
+        if has_aux:
+            # DualDPT: dim_in = 2 * embed_dim (because cat_token doubles token width).
+            dit_config["head_type"] = "dualdpt"
+            dit_config["head_output_dim"] = 2
+            dit_config["head_use_sky_head"] = False
+        else:
+            dit_config["head_type"] = "dpt"
+            dit_config["head_output_dim"] = state_dict[
+                '{}head.scratch.output_conv2.2.weight'.format(key_prefix)
+            ].shape[0]
+            dit_config["head_use_sky_head"] = (
+                '{}head.scratch.sky_output_conv2.0.weight'.format(key_prefix) in state_dict_keys
+            )
+
+        # out_layers: hard-coded per upstream YAML config (depth-aware default).
+        if depth >= 24:
+            # vitl: depths used vary between DA3-Large (DualDPT) and Mono/Metric (DPT).
+            if has_aux:
+                dit_config["out_layers"] = [11, 15, 19, 23]
+            else:
+                dit_config["out_layers"] = [4, 11, 17, 23]
+        else:
+            # vits/vitb: 12 blocks
+            dit_config["out_layers"] = [5, 7, 9, 11]
+
+        # Camera encoder/decoder presence (multi-view + pose path).
+        has_cam_enc = '{}cam_enc.token_norm.weight'.format(key_prefix) in state_dict_keys
+        has_cam_dec = '{}cam_dec.fc_t.weight'.format(key_prefix) in state_dict_keys
+        dit_config["has_cam_enc"] = has_cam_enc
+        dit_config["has_cam_dec"] = has_cam_dec
+        if has_cam_enc:
+            cam_enc_w = state_dict.get(
+                '{}cam_enc.pose_branch.fc2.weight'.format(key_prefix)
+            )
+            if cam_enc_w is not None:
+                dit_config["cam_dim_out"] = cam_enc_w.shape[0]
+        if has_cam_dec:
+            cam_dec_w = state_dict.get(
+                '{}cam_dec.fc_t.weight'.format(key_prefix)
+            )
+            if cam_dec_w is not None:
+                dit_config["cam_dec_dim_in"] = cam_dec_w.shape[1]
+        return dit_config
+
     if '{}layers.0.mlp.linear_fc2.weight'.format(key_prefix) in state_dict_keys: # Ernie Image
         dit_config = {}
         dit_config["image_model"] = "ernie"

comfy/model_management.py

@@ -542,8 +542,10 @@ try:
 except:
     pass
 
-if torch.cuda.is_available() and torch.backends.cudnn.is_available() and PerformanceFeature.AutoTune in args.fast:
-    torch.backends.cudnn.benchmark = True
+
+def set_cudnn_benchmark():
+    if torch.cuda.is_available() and torch.backends.cudnn.is_available():
+        torch.backends.cudnn.benchmark = PerformanceFeature.AutoTune in args.fast
 
 try:
     if torch_version_numeric >= (2, 5):
@@ -649,15 +651,19 @@ def free_pins(size, evict_active=False):
     return freed_total
 
 def ensure_pin_budget(size, evict_active=False):
-    shortfall = size + comfy.memory_management.RAM_CACHE_HEADROOM / 2 - psutil.virtual_memory().available
+    if args.high_ram:
+        return True
+    if args.fast_disk:
+        shortfall = TOTAL_PINNED_MEMORY + size - MAX_PINNED_MEMORY
+    else:
+        shortfall = size + max(comfy.memory_management.RAM_CACHE_HEADROOM / 2, 2048 * 1024 ** 2) - psutil.virtual_memory().available
     if shortfall <= 0:
         return True
 
     to_free = shortfall + PIN_PRESSURE_HYSTERESIS
     return free_pins(to_free, evict_active=evict_active) >= shortfall
 
-def ensure_pin_registerable(size, evict_active=False):
-    shortfall = TOTAL_PINNED_MEMORY + size - MAX_PINNED_MEMORY
+def free_registrations(shortfall, evict_active=True):
     if MAX_PINNED_MEMORY <= 0:
         return False
     if shortfall <= 0:
@@ -666,12 +672,22 @@ def ensure_pin_registerable(size, evict_active=False):
     shortfall += REGISTERABLE_PIN_HYSTERESIS
     for loaded_model in reversed(current_loaded_models):
         model = loaded_model.model
-        if model is not None and model.is_dynamic() and (evict_active or not model.model.dynamic_pins[model.load_device]["active"]):
+        if model is not None and model.is_dynamic() and not model.model.dynamic_pins[model.load_device]["active"]:
             shortfall -= model.unregister_inactive_pins(shortfall)
             if shortfall <= 0:
                 return True
+    if evict_active:
+        for loaded_model in current_loaded_models:
+            model = loaded_model.model
+            if model is not None and model.is_dynamic() and model.model.dynamic_pins[model.load_device]["active"]:
+                shortfall -= model.unregister_inactive_pins(shortfall)
+                if shortfall <= 0:
+                    return True
     return shortfall <= REGISTERABLE_PIN_HYSTERESIS
 
+def ensure_pin_registerable(size, evict_active=True):
+    return free_registrations(TOTAL_PINNED_MEMORY + size - MAX_PINNED_MEMORY, evict_active=evict_active)
+
 class LoadedModel:
     def __init__(self, model: ModelPatcher):
         self._set_model(model)
@@ -811,9 +827,9 @@ def free_memory(memory_required, device, keep_loaded=[], for_dynamic=False, pins
     for x in can_unload_sorted:
         i = x[-1]
         memory_to_free = 1e32
-        if current_loaded_models[i].model.is_dynamic() and (not DISABLE_SMART_MEMORY or device is None):
+        if not DISABLE_SMART_MEMORY or device is None:
             memory_to_free = 0 if device is None else memory_required - get_free_memory(device)
-            if for_dynamic:
+            if current_loaded_models[i].model.is_dynamic() and for_dynamic:
                 #don't actually unload dynamic models for the sake of other dynamic models
                 #as that works on-demand.
                 memory_required -= current_loaded_models[i].model.loaded_size()
@@ -825,6 +841,10 @@ def free_memory(memory_required, device, keep_loaded=[], for_dynamic=False, pins
     for i in sorted(unloaded_model, reverse=True):
         unloaded_models.append(current_loaded_models.pop(i))
 
+    if not for_dynamic and pins_required > 0:
+        ensure_pin_budget(pins_required)
+        ensure_pin_registerable(pins_required)
+
     if len(unloaded_model) > 0:
         soft_empty_cache()
     elif device is not None:
@@ -887,15 +907,19 @@ def load_models_gpu(models, memory_required=0, force_patch_weights=False, minimu
             model_to_unload.model_finalizer.detach()
 
     total_memory_required = {}
+    total_pins_required = {}
     for loaded_model in models_to_load:
         device = loaded_model.device
         total_memory_required[device] = total_memory_required.get(device, 0) + loaded_model.model_memory_required(device)
+        if not loaded_model.model.is_dynamic():
+            total_pins_required[device] = total_pins_required.get(device, 0) + loaded_model.model_memory()
 
     for device in total_memory_required:
         if device != torch.device("cpu"):
             free_memory(total_memory_required[device] * 1.1 + extra_mem,
                         device,
-                        for_dynamic=free_for_dynamic)
+                        for_dynamic=free_for_dynamic,
+                        pins_required=total_pins_required.get(device, 0))
 
     for device in total_memory_required:
         if device != torch.device("cpu"):
@@ -953,8 +977,6 @@ def loaded_models(only_currently_used=False):
 def cleanup_models_gc():
     do_gc = False
 
-    reset_cast_buffers()
-
     for i in range(len(current_loaded_models)):
         cur = current_loaded_models[i]
         if cur.is_dead():
@@ -1298,7 +1320,6 @@ STREAM_CAST_BUFFERS = {}
 LARGEST_CASTED_WEIGHT = (None, 0)
 STREAM_AIMDO_CAST_BUFFERS = {}
 LARGEST_AIMDO_CASTED_WEIGHT = (None, 0)
-STREAM_PIN_BUFFERS = {}
 
 DEFAULT_AIMDO_CAST_BUFFER_RESERVATION_SIZE = 16 * 1024 ** 3
 
@@ -1341,42 +1362,13 @@ def get_aimdo_cast_buffer(offload_stream, device):
         STREAM_AIMDO_CAST_BUFFERS[offload_stream] = cast_buffer
     return cast_buffer
 
-def get_pin_buffer(offload_stream):
-    pin_buffer = STREAM_PIN_BUFFERS.get(offload_stream, None)
-    if pin_buffer is None:
-        pin_buffer = comfy_aimdo.host_buffer.HostBuffer(0, 0, pinned_hostbuf_size(8 * 1024**3), mark_cold=False)
-        STREAM_PIN_BUFFERS[offload_stream] = pin_buffer
-    elif offload_stream is not None:
-        event = getattr(pin_buffer, "_comfy_event", None)
-        if event is not None:
-            event.synchronize()
-            delattr(pin_buffer, "_comfy_event")
-    return pin_buffer
-
-def resize_pin_buffer(pin_buffer, size):
-    global TOTAL_PINNED_MEMORY
-    old_size = pin_buffer.size
-    if size <= old_size:
-        return True
-    growth = size - old_size
-    comfy.memory_management.extra_ram_release(comfy.memory_management.RAM_CACHE_HEADROOM)
-    ensure_pin_budget(growth, evict_active=True)
-    ensure_pin_registerable(growth, evict_active=True)
-    try:
-        pin_buffer.extend(size=size, reallocate=True)
-    except RuntimeError:
-        return False
-    TOTAL_PINNED_MEMORY += pin_buffer.size - old_size
-    return True
-
 def reset_cast_buffers():
-    global TOTAL_PINNED_MEMORY
     global LARGEST_CASTED_WEIGHT
     global LARGEST_AIMDO_CASTED_WEIGHT
 
     LARGEST_CASTED_WEIGHT = (None, 0)
     LARGEST_AIMDO_CASTED_WEIGHT = (None, 0)
-    for offload_stream in set(STREAM_CAST_BUFFERS) | set(STREAM_AIMDO_CAST_BUFFERS) | set(STREAM_PIN_BUFFERS):
+    for offload_stream in set(STREAM_CAST_BUFFERS) | set(STREAM_AIMDO_CAST_BUFFERS):
         if offload_stream is not None:
             offload_stream.synchronize()
     synchronize()
@@ -1385,20 +1377,24 @@ def reset_cast_buffers():
         mmap_obj.bounce()
     DIRTY_MMAPS.clear()
 
-    for pin_buffer in STREAM_PIN_BUFFERS.values():
-        TOTAL_PINNED_MEMORY -= pin_buffer.size
-    TOTAL_PINNED_MEMORY = max(0, TOTAL_PINNED_MEMORY)
-
     for loaded_model in current_loaded_models:
         model = loaded_model.model
         if model is not None and model.is_dynamic():
-            model.model.dynamic_pins[model.load_device]["active"] = False
+            pin_state = model.model.dynamic_pins[model.load_device]
+
+            if pin_state["active"]:
+                *_, buckets = pin_state["weights"]
+                for size, bucket in list(buckets.items()):
+                    bucket[:] = [ entry for entry in bucket if entry[-1] is not None ]
+                    if not bucket:
+                        del buckets[size]
+
+            pin_state["active"] = False
             model.partially_unload_ram(1e30, subsets=[ "patches" ])
-            model.model.dynamic_pins[model.load_device]["patches"] = (comfy_aimdo.host_buffer.HostBuffer(0, 8 * 1024 * 1024, pinned_hostbuf_size(model.model_size())), [], [-1], [0])
+            model.model.dynamic_pins[model.load_device]["patches"] = (comfy_aimdo.host_buffer.HostBuffer(0, 8 * 1024 * 1024, pinned_hostbuf_size(model.model_size())), [], [-1], [0], [0], {})
 
     STREAM_CAST_BUFFERS.clear()
     STREAM_AIMDO_CAST_BUFFERS.clear()
-    STREAM_PIN_BUFFERS.clear()
     soft_empty_cache()
 
 def get_offload_stream(device):
@@ -1451,7 +1447,7 @@ def cast_to_gathered(tensors, r, non_blocking=False, stream=None, r2=None):
        if hasattr(wf_context, "as_context"):
            wf_context = wf_context.as_context(stream)
 
-    dest_views = comfy.memory_management.interpret_gathered_like(tensors, r)
+    dest_views = comfy.memory_management.interpret_gathered_like(tensors, r) if r is not None else [None] * len(tensors)
     dest2_views = comfy.memory_management.interpret_gathered_like(tensors, r2) if r2 is not None else None
     with wf_context:
         for tensor in tensors:
@@ -1463,9 +1459,10 @@ def cast_to_gathered(tensors, r, non_blocking=False, stream=None, r2=None):
                 continue
             storage = tensor._qdata.untyped_storage() if isinstance(tensor, comfy.quant_ops.QuantizedTensor) else tensor.untyped_storage()
             mark_mmap_dirty(storage)
-            dest_view.copy_(tensor, non_blocking=non_blocking)
+            if dest_view is not None:
+                dest_view.copy_(tensor, non_blocking=non_blocking)
             if dest2_view is not None:
-                dest2_view.copy_(dest_view, non_blocking=non_blocking)
+                dest2_view.copy_(tensor if dest_view is None else dest_view, non_blocking=non_blocking)
 
 
 def cast_to(weight, dtype=None, device=None, non_blocking=False, copy=False, stream=None, r=None):
@@ -1516,6 +1513,8 @@ if not args.disable_pinned_memory:
 PINNING_ALLOWED_TYPES = set(["Tensor", "Parameter", "QuantizedTensor"])
 
 def pinned_hostbuf_size(size):
+    if args.high_ram:
+        return max(0, int(size * 2))
     return max(0, int(min(size, MAX_PINNED_MEMORY) * 2))
 
 def discard_cuda_async_error():
@@ -1738,6 +1737,13 @@ def is_device_xpu(device):
 def is_device_cuda(device):
     return is_device_type(device, 'cuda')
 
+def set_torch_device(device):
+    """Set the current device for the given torch device. Supports CUDA and XPU."""
+    if is_device_cuda(device):
+        torch.cuda.set_device(device)
+    elif is_device_xpu(device):
+        torch.xpu.set_device(device)
+
 def is_directml_enabled():
     global directml_enabled
     if directml_enabled:

comfy/model_patcher.py

@@ -381,10 +381,11 @@ class ModelPatcher:
     def get_clone_model_override(self):
         return self.model, (self.backup, self.backup_buffers, self.object_patches_backup, self.pinned)
 
-    def clone(self, disable_dynamic=False, model_override=None):
+    def clone(self, disable_dynamic=False, model_override=None, force_deepcopy=False):
         class_ = self.__class__
-        if self.is_dynamic() and disable_dynamic:
-            class_ = ModelPatcher
+        if self.is_dynamic() and disable_dynamic or force_deepcopy:
+            if self.is_dynamic() and disable_dynamic:
+                class_ = ModelPatcher
             if model_override is None:
                 if self.cached_patcher_init is None:
                     raise RuntimeError("Cannot create non-dynamic delegate: cached_patcher_init is not initialized.")
@@ -1728,8 +1729,8 @@ class ModelPatcherDynamic(ModelPatcher):
         """
         if device not in self.model.dynamic_pins:
             self.model.dynamic_pins[device] = {
-                "weights": (comfy_aimdo.host_buffer.HostBuffer(0, 0, 0), [], [-1], [0]),
-                "patches": (comfy_aimdo.host_buffer.HostBuffer(0, 0, 0), [], [-1], [0]),
+                "weights": (comfy_aimdo.host_buffer.HostBuffer(0, 0, 0), [], [-1], [0], [0], {}),
+                "patches": (comfy_aimdo.host_buffer.HostBuffer(0, 0, 0), [], [-1], [0], [0], {}),
                 "hostbufs_initialized": False,
                 "failed": False,
                 "active": False,
@@ -1806,8 +1807,8 @@ class ModelPatcherDynamic(ModelPatcher):
             pin_state = self.model.dynamic_pins[self.load_device]
             if not pin_state["hostbufs_initialized"]:
                 hostbuf_size = comfy.model_management.pinned_hostbuf_size(self.model_size())
-                pin_state["weights"] = (comfy_aimdo.host_buffer.HostBuffer(0, 64 * 1024 * 1024, hostbuf_size), [], [-1], [0])
-                pin_state["patches"] = (comfy_aimdo.host_buffer.HostBuffer(0, 8 * 1024 * 1024, hostbuf_size), [], [-1], [0])
+                pin_state["weights"] = (comfy_aimdo.host_buffer.HostBuffer(0, 64 * 1024 * 1024, hostbuf_size), [], [-1], [0], [0], {})
+                pin_state["patches"] = (comfy_aimdo.host_buffer.HostBuffer(0, 8 * 1024 * 1024, hostbuf_size), [], [-1], [0], [0], {})
                 pin_state["hostbufs_initialized"] = True
             pin_state["failed"] = False
             pin_state["active"] = True
@@ -1949,18 +1950,16 @@ class ModelPatcherDynamic(ModelPatcher):
         return freed
 
     def loaded_ram_size(self):
-        return (self.model.dynamic_pins[self.load_device]["weights"][0].size +
-                self.model.dynamic_pins[self.load_device]["patches"][0].size)
+        return (self.model.dynamic_pins[self.load_device]["weights"][0].size)
 
     def pinned_memory_size(self):
-        return (self.model.dynamic_pins[self.load_device]["weights"][3][0] +
-                self.model.dynamic_pins[self.load_device]["patches"][3][0])
+        return (self.model.dynamic_pins[self.load_device]["weights"][3][0])
 
     def unregister_inactive_pins(self, ram_to_unload, subsets=[ "weights", "patches" ]):
         freed = 0
         pin_state = self.model.dynamic_pins[self.load_device]
         for subset in subsets:
-            hostbuf, stack, stack_split, pinned_size = pin_state[subset]
+            hostbuf, stack, stack_split, pinned_size, *_ = pin_state[subset]
             split = stack_split[0]
             while split >= 0:
                 module, offset = stack[split]
@@ -1985,10 +1984,12 @@ class ModelPatcherDynamic(ModelPatcher):
         freed = 0
         pin_state = self.model.dynamic_pins[self.load_device]
         for subset in subsets:
-            hostbuf, stack, stack_split, pinned_size = pin_state[subset]
+            hostbuf, stack, stack_split, pinned_size, *_ = pin_state[subset]
             while len(stack) > 0:
                 module, offset = stack.pop()
                 size = module._pin.numel() * module._pin.element_size()
+                module._pin_balancer_entry[-1] = None
+                del module._pin_balancer_entry
                 del module._pin
                 hostbuf.truncate(offset, do_unregister=module._pin_registered)
                 stack_split[0] = min(stack_split[0], len(stack) - 1)

comfy/model_prefetch.py

@@ -1,4 +1,5 @@
 import comfy_aimdo.model_vbar
+import comfy.memory_management
 import comfy.model_management
 import comfy.ops
 
@@ -50,7 +51,17 @@ def prefetch_queue_pop(queue, device, module):
             if hasattr(s, "_v"):
                 comfy_modules.append(s)
 
+        registerable_size = 0
+        for s in comfy_modules:
+            registerable_size += comfy.memory_management.vram_aligned_size([s.weight, s.bias])
+            for param_key in ("weight", "bias"):
+                lowvram_fn = getattr(s, param_key + "_lowvram_function", None)
+                if lowvram_fn is not None:
+                    registerable_size += lowvram_fn.memory_required()
+
         offload_stream = comfy.ops.cast_modules_with_vbar(comfy_modules, None, device, None, True)
+        if not comfy.model_management.args.fast_disk:
+            comfy.model_management.ensure_pin_registerable(registerable_size)
         comfy.model_management.sync_stream(device, offload_stream)
         queue[0] = (offload_stream, (prefetch, comfy_modules))
 

comfy/multigpu.py

@@ -17,7 +17,7 @@ class MultiGPUThreadPool:
     """Persistent thread pool for multi-GPU work distribution.
 
     Maintains one worker thread per extra GPU device. Each thread calls
-    torch.cuda.set_device() once at startup so that compiled kernel caches
+    set_torch_device() once at startup so that compiled kernel caches
     (inductor/triton) stay warm across diffusion steps.
     """
 
@@ -37,7 +37,7 @@ class MultiGPUThreadPool:
 
     def _worker_loop(self, device: torch.device, work_q: queue.Queue, result_q: queue.Queue):
         try:
-            torch.cuda.set_device(device)
+            comfy.model_management.set_torch_device(device)
         except Exception as e:
             logging.error(f"MultiGPUThreadPool: failed to set device {device}: {e}")
             while True:
@@ -54,6 +54,8 @@ class MultiGPUThreadPool:
             try:
                 result = fn(*args, **kwargs)
                 result_q.put((result, None))
+            except comfy.model_management.InterruptProcessingException as e:
+                result_q.put((None, e))
             except Exception as e:
                 result_q.put((None, e))
 

comfy/ops.py

@@ -76,8 +76,6 @@ except:
 
 cast_to = comfy.model_management.cast_to #TODO: remove once no more references
 
-STREAM_PIN_BUFFER_HEADROOM = 8 * 1024 * 1024
-
 def cast_to_input(weight, input, non_blocking=False, copy=True):
     return comfy.model_management.cast_to(weight, input.dtype, input.device, non_blocking=non_blocking, copy=copy)
 
@@ -94,9 +92,6 @@ def cast_modules_with_vbar(comfy_modules, dtype, device, bias_dtype, non_blockin
     offload_stream = None
     cast_buffer = None
     cast_buffer_offset = 0
-    stream_pin_hostbuf = None
-    stream_pin_offset = 0
-    stream_pin_queue = []
 
     def ensure_offload_stream(module, required_size, check_largest):
         nonlocal offload_stream
@@ -130,22 +125,6 @@ def cast_modules_with_vbar(comfy_modules, dtype, device, bias_dtype, non_blockin
         cast_buffer_offset += buffer_size
         return buffer
 
-    def get_stream_pin_buffer_offset(buffer_size):
-        nonlocal stream_pin_hostbuf
-        nonlocal stream_pin_offset
-
-        if buffer_size == 0 or offload_stream is None:
-            return None
-
-        if stream_pin_hostbuf is None:
-            stream_pin_hostbuf = comfy.model_management.get_pin_buffer(offload_stream)
-            if stream_pin_hostbuf is None:
-                return None
-
-        offset = stream_pin_offset
-        stream_pin_offset += buffer_size
-        return offset
-
     for s in comfy_modules:
         signature = comfy_aimdo.model_vbar.vbar_fault(s._v)
         resident = comfy_aimdo.model_vbar.vbar_signature_compare(signature, s._v_signature)
@@ -184,33 +163,27 @@ def cast_modules_with_vbar(comfy_modules, dtype, device, bias_dtype, non_blockin
         if xfer_dest is None:
             xfer_dest = get_cast_buffer(dest_size)
 
-        def cast_maybe_lowvram_patch(xfer_source, xfer_dest, stream):
+        def cast_maybe_lowvram_patch(xfer_source, xfer_dest, stream, xfer_dest2=None):
             if xfer_source is not None:
                 if getattr(xfer_source, "is_lowvram_patch", False):
-                    xfer_source.prepare(xfer_dest, stream, copy=True, commit=False)
-                else:
-                    comfy.model_management.cast_to_gathered(xfer_source, xfer_dest, non_blocking=non_blocking, stream=stream)
+                    if xfer_dest is not None:
+                        xfer_source.prepare(xfer_dest, stream, copy=True, commit=False)
+                        xfer_source = [ xfer_dest ]
+                        xfer_dest = xfer_dest2
+                        xfer_dest2 = None
+                    elif xfer_dest2 is not None:
+                        xfer_source.prepare(xfer_dest2, stream, copy=True, commit=False)
+                        return
+                comfy.model_management.cast_to_gathered(xfer_source, xfer_dest, non_blocking=non_blocking, stream=stream, r2=xfer_dest2)
 
         def handle_pin(m, pin, source, dest, subset="weights", size=None):
             if pin is not None:
                 cast_maybe_lowvram_patch([pin], dest, offload_stream)
                 return
-            if signature is None:
+            if signature is None or args.high_ram:
                 comfy.pinned_memory.pin_memory(m, subset=subset, size=size)
                 pin = comfy.pinned_memory.get_pin(m, subset=subset)
-                if pin is not None:
-                    if isinstance(source, list):
-                        comfy.model_management.cast_to_gathered(source, pin, non_blocking=non_blocking, stream=offload_stream, r2=dest)
-                    else:
-                        cast_maybe_lowvram_patch(source, pin, None)
-                        cast_maybe_lowvram_patch([ pin ], dest, offload_stream)
-                    return
-            if pin is None:
-                pin_offset = get_stream_pin_buffer_offset(size)
-                if pin_offset is not None:
-                    stream_pin_queue.append((source, pin_offset, size, dest))
-                    return
-            cast_maybe_lowvram_patch(source, dest, offload_stream)
+            cast_maybe_lowvram_patch(source, pin, offload_stream, xfer_dest2=dest)
 
         handle_pin(s, pin, xfer_source, xfer_dest, size=dest_size)
 
@@ -232,23 +205,6 @@ def cast_modules_with_vbar(comfy_modules, dtype, device, bias_dtype, non_blockin
         prefetch["needs_cast"] = needs_cast
         s._prefetch = prefetch
 
-    if stream_pin_offset > 0:
-        if stream_pin_hostbuf.size < stream_pin_offset:
-            if not comfy.model_management.resize_pin_buffer(stream_pin_hostbuf, stream_pin_offset + STREAM_PIN_BUFFER_HEADROOM):
-                for xfer_source, _, _, xfer_dest in stream_pin_queue:
-                    cast_maybe_lowvram_patch(xfer_source, xfer_dest, offload_stream)
-                return offload_stream
-        stream_pin_tensor = comfy_aimdo.torch.hostbuf_to_tensor(stream_pin_hostbuf)
-        stream_pin_tensor.untyped_storage()._comfy_hostbuf = stream_pin_hostbuf
-        for xfer_source, pin_offset, pin_size, xfer_dest in stream_pin_queue:
-            pin = stream_pin_tensor[pin_offset:pin_offset + pin_size]
-            if isinstance(xfer_source, list):
-                comfy.model_management.cast_to_gathered(xfer_source, pin, non_blocking=non_blocking, stream=offload_stream, r2=xfer_dest)
-            else:
-                cast_maybe_lowvram_patch(xfer_source, pin, None)
-                comfy.model_management.cast_to_gathered([ pin ], xfer_dest, non_blocking=non_blocking, stream=offload_stream)
-        stream_pin_hostbuf._comfy_event = offload_stream.record_event()
-
     return offload_stream
 
 
@@ -343,21 +299,21 @@ def cast_bias_weight(s, input=None, dtype=None, device=None, bias_dtype=None, of
 
     non_blocking = comfy.model_management.device_supports_non_blocking(device)
 
-    if hasattr(s, "_v"):
+    if hasattr(s, "_v") and comfy.model_management.is_device_cpu(device):
 
         #vbar doesn't support CPU weights, but some custom nodes have weird paths
         #that might switch the layer to the CPU and expect it to work. We have to take
         #a clone conservatively as we are mmapped and some SFT files are packed misaligned
         #If you are a custom node author reading this, please move your layer to the GPU
         #or declare your ModelPatcher as CPU in the first place.
-        if comfy.model_management.is_device_cpu(device):
-            materialize_meta_param(s, ["weight", "bias"])
-            weight = s.weight.to(dtype=dtype, copy=True)
-            if isinstance(weight, QuantizedTensor):
-                weight = weight.dequantize()
-            bias = s.bias.to(dtype=bias_dtype, copy=True) if s.bias is not None else None
-            return format_return((weight, bias, (None, None, None)), offloadable)
+        materialize_meta_param(s, ["weight", "bias"])
+        weight = s.weight.to(dtype=dtype, copy=True)
+        if isinstance(weight, QuantizedTensor):
+            weight = weight.dequantize()
+        bias = s.bias.to(dtype=bias_dtype, copy=True) if s.bias is not None else None
+        return format_return((weight, bias, (None, None, None)), offloadable)
 
+    elif hasattr(s, "_v") and s.weight.device != device:
         prefetched = hasattr(s, "_prefetch")
         offload_stream = None
         offload_device = None

comfy/pinned_memory.py

@@ -1,17 +1,55 @@
+import bisect
+
 import comfy.model_management
 import comfy.memory_management
+import comfy.utils
 import comfy_aimdo.host_buffer
 import comfy_aimdo.torch
 import torch
 
 from comfy.cli_args import args
 
+def _add_to_bucket(module, buckets, size, priority):
+    bucket = buckets.setdefault(size, [])
+    entry = [-priority, 0, module]
+    entry[1] = id(entry)
+    bisect.insort(bucket, entry)
+    module._pin_balancer_entry = entry
+
+def _steal_pin(module, stack, buckets, size, priority):
+    bucket = buckets.get(size)
+    if bucket is None:
+        return False
+
+    while bucket and bucket[-1][-1] is None:
+        bucket.pop()
+    if not bucket:
+        del buckets[size]
+        return False
+
+    if priority <= -bucket[-1][0]:
+        return False
+
+    *_, victim = bucket.pop()
+    module._pin = victim._pin
+    module._pin_registered = victim._pin_registered
+    module._pin_stack_index = victim._pin_stack_index
+    stack[module._pin_stack_index] = (module, stack[module._pin_stack_index][1])
+
+    victim._pin_registered = False
+    del victim._pin
+    del victim._pin_stack_index
+    del victim._pin_balancer_entry
+
+    _add_to_bucket(module, buckets, size, priority)
+    return True
+
 def get_pin(module, subset="weights"):
     pin = getattr(module, "_pin", None)
     if pin is None or module._pin_registered or args.disable_pinned_memory:
         return pin
 
-    _, _, stack_split, pinned_size = module._pin_state[subset]
+    _, _, stack_split, pinned_size, *_ = module._pin_state[subset]
     size = pin.nbytes
     comfy.model_management.ensure_pin_registerable(size)
 
@@ -31,33 +69,51 @@ def pin_memory(module, subset="weights", size=None):
         return
 
     pin = get_pin(module, subset)
-    if pin is not None or pin_state["failed"]:
+    if pin is not None:
         return
 
-    hostbuf, stack, stack_split, pinned_size = pin_state[subset]
+    hostbuf, stack, stack_split, pinned_size, counter, buckets = pin_state[subset]
     if size is None:
         size = comfy.memory_management.vram_aligned_size([ module.weight, module.bias ])
     offset = hostbuf.size
-    registerable_size = size + max(0, hostbuf.size - pinned_size[0])
+    registerable_size = size
+    priority = getattr(module, "_pin_balancer_priority", None)
+
+    if priority is None:
+        priority = comfy.utils.bit_reverse_range(counter[0], 16)
+        counter[0] += 1
+        module._pin_balancer_priority = priority
 
     comfy.memory_management.extra_ram_release(comfy.memory_management.RAM_CACHE_HEADROOM)
     if (not comfy.model_management.ensure_pin_budget(size) or
         not comfy.model_management.ensure_pin_registerable(registerable_size)):
-        pin_state["failed"] = True
-        return False
+        return _steal_pin(module, stack, buckets, size, priority)
 
+    extended = False
     try:
-        hostbuf.extend(size=size)
+        hostbuf.extend(size=size, register=False)
+        extended = True
+        pin = comfy_aimdo.torch.hostbuf_to_tensor(hostbuf)[offset:offset + size]
+        pin.untyped_storage()._comfy_hostbuf = hostbuf
+        if torch.cuda.cudart().cudaHostRegister(pin.data_ptr(), size, 1) != 0:
+            comfy.model_management.discard_cuda_async_error()
+            comfy.model_management.free_registrations(size)
+            if torch.cuda.cudart().cudaHostRegister(pin.data_ptr(), size, 1) != 0:
+                comfy.model_management.discard_cuda_async_error()
+                del pin
+                hostbuf.truncate(offset, do_unregister=False)
+                return _steal_pin(module, stack, buckets, size, priority)
     except RuntimeError:
-        pin_state["failed"] = True
-        return False
+        if extended:
+            hostbuf.truncate(offset, do_unregister=False)
+        return _steal_pin(module, stack, buckets, size, priority)
 
-    module._pin = comfy_aimdo.torch.hostbuf_to_tensor(hostbuf)[offset:offset + size]
-    module._pin.untyped_storage()._comfy_hostbuf = hostbuf
+    module._pin = pin
     stack.append((module, offset))
     module._pin_registered = True
     module._pin_stack_index = len(stack) - 1
     stack_split[0] = max(stack_split[0], module._pin_stack_index)
     comfy.model_management.TOTAL_PINNED_MEMORY += size
     pinned_size[0] += size
+    _add_to_bucket(module, buckets, size, priority)
     return True

comfy/samplers.py

@@ -464,10 +464,7 @@ def _calc_cond_batch_multigpu(model: BaseModel, conds: list[list[dict]], x_in: t
 
     def _handle_batch(device: torch.device, batch_tuple: tuple[comfy.hooks.HookGroup, tuple], results: list[thread_result]):
         try:
-            # TODO: non-NVIDIA support -- guard with `if device.type == "cuda":` once
-            # we extend multigpu QA beyond CUDA. Unconditional call crashes on
-            # XPU/NPU/MPS/CPU/DirectML backends.
-            torch.cuda.set_device(device)
+            comfy.model_management.set_torch_device(device)
             model_current: BaseModel = model_options["multigpu_clones"][device].model
             # run every hooked_to_run separately
             with torch.no_grad():

comfy/sd.py

@@ -16,6 +16,7 @@ import comfy.ldm.cosmos.vae
 import comfy.ldm.wan.vae
 import comfy.ldm.wan.vae2_2
 import comfy.ldm.hunyuan3d.vae
+import comfy.ldm.triposplat.vae
 import comfy.ldm.ace.vae.music_dcae_pipeline
 import comfy.ldm.cogvideo.vae
 import comfy.ldm.hunyuan_video.vae
@@ -57,6 +58,7 @@ import comfy.text_encoders.omnigen2
 import comfy.text_encoders.qwen_image
 import comfy.text_encoders.hunyuan_image
 import comfy.text_encoders.z_image
+import comfy.text_encoders.ideogram4
 import comfy.text_encoders.ovis
 import comfy.text_encoders.kandinsky5
 import comfy.text_encoders.jina_clip_2
@@ -65,6 +67,8 @@ import comfy.text_encoders.anima
 import comfy.text_encoders.ace15
 import comfy.text_encoders.longcat_image
 import comfy.text_encoders.qwen35
+import comfy.text_encoders.qwen3vl
+import comfy.text_encoders.boogu
 import comfy.text_encoders.ernie
 import comfy.text_encoders.gemma4
 import comfy.text_encoders.cogvideo
@@ -894,6 +898,16 @@ class VAE:
                 #Force cast it for --disable-dynamic-vram users until there is a true core fix.
                 if not comfy.memory_management.aimdo_enabled:
                     self.disable_offload = True
+            elif "gs.base_offset_scale" in sd and "octree.out_proj.weight" in sd:  # TripoSplat octree gaussian decoder
+                self.first_stage_model = comfy.ldm.triposplat.vae.OctreeGaussianDecoder()
+                self.latent_channels = 16
+                self.latent_dim = 1
+                self.working_dtypes = [torch.float16, torch.bfloat16, torch.float32]
+                # The generic VAE.encode/decode path isn't used: VAEDecodeTripoSplat calls the gaussian
+                # decoder directly (structured GaussianSplat objects, not a tensor and reserves VRAM itself from num_gaussians.
+                def _no_generic_io(*args, **kwargs):
+                    raise RuntimeError("TripoSplat gaussian decoder: use the 'TripoSplat Decode' (VAEDecodeTripoSplat)")
+                self.memory_used_encode = self.memory_used_decode = _no_generic_io
             else:
                 logging.warning("WARNING: No VAE weights detected, VAE not initalized.")
                 self.first_stage_model = None
@@ -1297,6 +1311,8 @@ class CLIPType(Enum):
     COGVIDEOX = 27
     LENS = 28
     PIXELDIT = 29
+    IDEOGRAM4 = 30
+    BOOGU = 31
 
 
 
@@ -1350,6 +1366,8 @@ class TEModel(Enum):
     GEMMA_4_31B = 31
     T5_GEMMA = 32
     GPT_OSS_20B = 33
+    QWEN3VL_4B = 34
+    QWEN3VL_8B = 35
 
 
 def detect_te_model(sd):
@@ -1411,6 +1429,8 @@ def detect_te_model(sd):
         if weight.shape[0] == 5120:
             return TEModel.QWEN35_27B
         return TEModel.QWEN35_2B
+    if "model.visual.deepstack_merger_list.0.norm.weight" in sd:  # DeepStack is unique to Qwen3-VL
+        return TEModel.QWEN3VL_4B if sd["model.visual.merger.linear_fc2.weight"].shape[0] == 2560 else TEModel.QWEN3VL_8B
     if "model.layers.0.post_attention_layernorm.weight" in sd:
         weight = sd['model.layers.0.post_attention_layernorm.weight']
         if 'model.layers.0.self_attn.q_norm.weight' in sd:
@@ -1595,8 +1615,12 @@ def load_text_encoder_state_dicts(state_dicts=[], embedding_directory=None, clip
             clip_target.clip = comfy.text_encoders.ovis.te(**llama_detect(clip_data))
             clip_target.tokenizer = comfy.text_encoders.ovis.OvisTokenizer
         elif te_model == TEModel.QWEN3_8B:
-            clip_target.clip = comfy.text_encoders.flux.klein_te(**llama_detect(clip_data), model_type="qwen3_8b")
-            clip_target.tokenizer = comfy.text_encoders.flux.KleinTokenizer8B
+            if clip_type == CLIPType.IDEOGRAM4:
+                clip_target.clip = comfy.text_encoders.ideogram4.te(**llama_detect(clip_data))
+                clip_target.tokenizer = comfy.text_encoders.ideogram4.Ideogram4Tokenizer
+            else:
+                clip_target.clip = comfy.text_encoders.flux.klein_te(**llama_detect(clip_data), model_type="qwen3_8b")
+                clip_target.tokenizer = comfy.text_encoders.flux.KleinTokenizer8B
         elif te_model == TEModel.JINA_CLIP_2:
             clip_target.clip = comfy.text_encoders.jina_clip_2.JinaClip2TextModelWrapper
             clip_target.tokenizer = comfy.text_encoders.jina_clip_2.JinaClip2TokenizerWrapper
@@ -1605,6 +1629,24 @@ def load_text_encoder_state_dicts(state_dicts=[], embedding_directory=None, clip
             qwen35_type = {TEModel.QWEN35_08B: "qwen35_08b", TEModel.QWEN35_2B: "qwen35_2b", TEModel.QWEN35_4B: "qwen35_4b", TEModel.QWEN35_9B: "qwen35_9b", TEModel.QWEN35_27B: "qwen35_27b"}[te_model]
             clip_target.clip = comfy.text_encoders.qwen35.te(**llama_detect(clip_data), model_type=qwen35_type)
             clip_target.tokenizer = comfy.text_encoders.qwen35.tokenizer(model_type=qwen35_type)
+        elif te_model in (TEModel.QWEN3VL_4B, TEModel.QWEN3VL_8B):
+            if clip_type == CLIPType.IDEOGRAM4 and te_model == TEModel.QWEN3VL_8B:  # Ideogram4 reuses the full Qwen3-VL-8B (13-layer tap for conditioning + multimodal generate).
+                clip_data[0] = comfy.utils.state_dict_prefix_replace(clip_data[0], {"model.language_model.": "model.", "model.visual.": "visual.", "lm_head.": "model.lm_head."})
+                clip_target.clip = comfy.text_encoders.ideogram4.te_qwen3vl(**llama_detect(clip_data))
+                clip_target.tokenizer = comfy.text_encoders.ideogram4.Ideogram4Qwen3VLTokenizer
+            elif clip_type == CLIPType.BOOGU and te_model == TEModel.QWEN3VL_8B:  # Boogu-Image: full Qwen3-VL-8B, last hidden state, no-think template.
+                clip_data[0] = comfy.utils.state_dict_prefix_replace(clip_data[0], {"model.language_model.": "model.", "model.visual.": "visual.", "lm_head.": "model.lm_head."})
+                clip_target.clip = comfy.text_encoders.boogu.te(**llama_detect(clip_data))
+                clip_target.tokenizer = comfy.text_encoders.boogu.BooguTokenizer
+            elif clip_type in (CLIPType.FLUX, CLIPType.FLUX2):  # Flux2 Klein reuses the Qwen3-VL LM (3-layer tap -> 12288); visual unused.
+                klein_model_type = "qwen3_8b" if te_model == TEModel.QWEN3VL_8B else "qwen3_4b"
+                clip_target.clip = comfy.text_encoders.flux.klein_te(**llama_detect(clip_data), model_type=klein_model_type)
+                clip_target.tokenizer = comfy.text_encoders.flux.KleinTokenizer8B if te_model == TEModel.QWEN3VL_8B else comfy.text_encoders.flux.KleinTokenizer
+            else:
+                clip_data[0] = comfy.utils.state_dict_prefix_replace(clip_data[0], {"model.language_model.": "model.", "model.visual.": "visual.", "lm_head.": "model.lm_head."})
+                qwen3vl_type = {TEModel.QWEN3VL_4B: "qwen3vl_4b", TEModel.QWEN3VL_8B: "qwen3vl_8b"}[te_model]
+                clip_target.clip = comfy.text_encoders.qwen3vl.te(**llama_detect(clip_data), model_type=qwen3vl_type)
+                clip_target.tokenizer = comfy.text_encoders.qwen3vl.tokenizer(model_type=qwen3vl_type)
         elif te_model == TEModel.QWEN3_06B:
             clip_target.clip = comfy.text_encoders.anima.te(**llama_detect(clip_data))
             clip_target.tokenizer = comfy.text_encoders.anima.AnimaTokenizer

comfy/supported_models.py

@@ -24,6 +24,8 @@ import comfy.text_encoders.qwen_image
 import comfy.text_encoders.hunyuan_image
 import comfy.text_encoders.kandinsky5
 import comfy.text_encoders.z_image
+import comfy.text_encoders.ideogram4
+import comfy.text_encoders.boogu
 import comfy.text_encoders.anima
 import comfy.text_encoders.ace15
 import comfy.text_encoders.longcat_image
@@ -1449,6 +1451,17 @@ class WAN21_SCAIL(WAN21_T2V):
         out = model_base.WAN21_SCAIL(self, image_to_video=False, device=device)
         return out
 
+
+class WAN21_SCAIL2(WAN21_T2V):
+    unet_config = {
+        "image_model": "wan2.1",
+        "model_type": "scail2",
+    }
+
+    def get_model(self, state_dict, prefix="", device=None):
+        out = model_base.WAN21_SCAIL2(self, image_to_video=False, device=device)
+        return out
+
 class WAN22_WanDancer(WAN21_T2V):
     unet_config = {
         "image_model": "wan2.1",
@@ -1538,6 +1551,30 @@ class Hunyuan3Dv2mini(Hunyuan3Dv2):
 
     latent_format = latent_formats.Hunyuan3Dv2mini
 
+class TripoSplat(supported_models_base.BASE):
+    # Image -> 3D gaussian splat flow denoiser
+    unet_config = {
+        "image_model": "triposplat",
+    }
+
+    unet_extra_config = {}
+
+    sampling_settings = {
+        "shift": 3.0,
+    }
+
+    memory_usage_factor = 0.6
+
+    latent_format = latent_formats.TripoSplat
+
+    supported_inference_dtypes = [torch.float16, torch.bfloat16, torch.float32]
+
+    def get_model(self, state_dict, prefix="", device=None):
+        return model_base.TripoSplat(self, device=device)
+
+    def clip_target(self, state_dict={}):
+        return None
+
 class HiDream(supported_models_base.BASE):
     unet_config = {
         "image_model": "hidream",
@@ -1722,6 +1759,65 @@ class Omnigen2(supported_models_base.BASE):
         hunyuan_detect = comfy.text_encoders.hunyuan_video.llama_detect(state_dict, "{}qwen25_3b.transformer.".format(pref))
         return supported_models_base.ClipTarget(comfy.text_encoders.omnigen2.Omnigen2Tokenizer, comfy.text_encoders.omnigen2.te(**hunyuan_detect))
 
+class Boogu(Omnigen2):
+    unet_config = {
+        "image_model": "boogu",
+    }
+
+    sampling_settings = {
+        "multiplier": 1.0,
+        "shift": 3.16,
+    }
+
+    memory_usage_factor = 2.15
+
+    def get_model(self, state_dict, prefix="", device=None):
+        out = model_base.Boogu(self, device=device)
+        return out
+
+    def clip_target(self, state_dict={}):
+        pref = self.text_encoder_key_prefix[0]
+        hunyuan_detect = comfy.text_encoders.hunyuan_video.llama_detect(state_dict, "{}qwen3vl_8b.transformer.".format(pref))
+        return supported_models_base.ClipTarget(comfy.text_encoders.boogu.BooguTokenizer, comfy.text_encoders.boogu.te(**hunyuan_detect))
+
+class Ideogram4(supported_models_base.BASE):
+    unet_config = {
+        "image_model": "ideogram4",
+    }
+
+    sampling_settings = {
+        "multiplier": 1.0,
+        "shift": 1.0,
+    }
+
+    memory_usage_factor = 11.6
+
+    unet_extra_config = {
+        "num_attention_heads": 18,
+        "attention_head_dim": 256,
+        "intermediate_size": 12288,
+        "adaln_dim": 512,
+        "llm_features_dim": 53248,
+        "rope_theta": 5000000,
+        "mrope_section": [24, 20, 20],
+        "norm_eps": 1e-5,
+    }
+    latent_format = latent_formats.Flux2
+
+    supported_inference_dtypes = [torch.bfloat16, torch.float32]
+
+    vae_key_prefix = ["vae."]
+    text_encoder_key_prefix = ["text_encoders."]
+
+    def get_model(self, state_dict, prefix="", device=None):
+        out = model_base.Ideogram4(self, device=device)
+        return out
+
+    def clip_target(self, state_dict={}):
+        pref = self.text_encoder_key_prefix[0]
+        hunyuan_detect = comfy.text_encoders.hunyuan_video.llama_detect(state_dict, "{}qwen3vl_8b.transformer.".format(pref))
+        return supported_models_base.ClipTarget(comfy.text_encoders.ideogram4.Ideogram4Tokenizer, comfy.text_encoders.ideogram4.te(**hunyuan_detect))
+
 class QwenImage(supported_models_base.BASE):
     unet_config = {
         "image_model": "qwen_image",
@@ -1982,6 +2078,23 @@ class RT_DETR_v4(supported_models_base.BASE):
         return None
 
 
+class DepthAnything3(supported_models_base.BASE):
+    unet_config = {
+        "image_model": "DepthAnything3",
+    }
+
+    # Mono path: no num_heads / num_head_channels needed.
+    unet_extra_config = {}
+
+    supported_inference_dtypes = [torch.float16, torch.bfloat16, torch.float32]
+
+    def get_model(self, state_dict, prefix="", device=None):
+        return model_base.DepthAnything3(self, device=device)
+
+    def clip_target(self, state_dict={}):
+        return None
+
+
 class ErnieImage(supported_models_base.BASE):
     unet_config = {
         "image_model": "ernie",
@@ -2196,10 +2309,12 @@ models = [
     WAN22_Animate,
     WAN21_FlowRVS,
     WAN21_SCAIL,
+    WAN21_SCAIL2,
     WAN22_WanDancer,
     Hunyuan3Dv2mini,
     Hunyuan3Dv2,
     Hunyuan3Dv2_1,
+    TripoSplat,
     HiDream,
     HiDreamO1,
     Chroma,
@@ -2207,7 +2322,9 @@ models = [
     ACEStep,
     ACEStep15,
     Omnigen2,
+    Boogu,
     QwenImage,
+    Ideogram4,
     Flux2,
     Lens,
     Kandinsky5Image,
@@ -2221,4 +2338,5 @@ models = [
     CogVideoX_I2V,
     CogVideoX_T2V,
     SVD_img2vid,
+    DepthAnything3,
 ]

comfy/text_encoders/boogu.py

@@ -0,0 +1,58 @@
+"""Boogu-Image text encoder: full Qwen3-VL-8B, last hidden state (4096-dim).
+
+Boogu uses the final hidden state of Qwen3-VL as the per-token instruction feature
+(num_instruction_feature_layers=1, reduce_type=mean -> just the last layer).
+The model itself is the standard Qwen3-VL TE, only the chat template differs
+(a fixed system prompt and no <think> block).
+"""
+
+import comfy.text_encoders.qwen3vl
+from comfy import sd1_clip
+
+
+# System prompts from the reference pipeline (pipeline_boogu.py).
+# T2I (non-empty instruction, no image) uses the helpful-assistant prompt
+# everything else (the CFG negative / "drop" condition, and any image case) uses the TI2I "describe" prompt.
+BOOGU_T2I_SYSTEM = "You are a helpful assistant that generates high-quality images based on user instructions. The instructions are as follows."
+BOOGU_DROP_SYSTEM = "Describe the key features of the input image (color, shape, size, texture, objects, background), then explain how the user's text instruction should alter or modify the image. Generate a new image that meets the user's requirements while maintaining consistency with the original input where appropriate."
+
+
+class BooguTokenizer(comfy.text_encoders.qwen3vl.Qwen3VLTokenizer):
+    def __init__(self, embedding_directory=None, tokenizer_data={}):
+        super().__init__(embedding_directory=embedding_directory, tokenizer_data=tokenizer_data, model_type="qwen3vl_8b")
+        # apply_chat_template without add_generation_prompt
+        self.llama_template = "<|im_start|>system\n" + BOOGU_T2I_SYSTEM + "<|im_end|>\n<|im_start|>user\n{}<|im_end|>\n"
+        self.llama_template_images = "<|im_start|>system\n" + BOOGU_DROP_SYSTEM + "<|im_end|>\n<|im_start|>user\n<|vision_start|><|image_pad|><|vision_end|>{}<|im_end|>\n"
+        # Reference SYSTEM_PROMPT_DROP: used for the empty negative/uncond instruction.
+        self.llama_template_drop = "<|im_start|>system\n" + BOOGU_DROP_SYSTEM + "<|im_end|>\n<|im_start|>user\n{}<|im_end|>\n"
+
+    def tokenize_with_weights(self, text, return_word_ids=False, llama_template=None, images=[], prevent_empty_text=False, thinking=True, **kwargs):
+        if llama_template is None and len(images) == 0 and text.strip() == "":
+            llama_template = self.llama_template_drop
+        # Boogu conditions on the no-think template; thinking=True drops the empty <think> block qwen3vl adds by default.
+        return super().tokenize_with_weights(text, return_word_ids=return_word_ids, llama_template=llama_template, images=images, prevent_empty_text=prevent_empty_text, thinking=thinking, **kwargs)
+
+
+class BooguQwen3VLClipModel(comfy.text_encoders.qwen3vl.Qwen3VLClipModel):
+    def __init__(self, device="cpu", dtype=None, attention_mask=True, model_options={}, model_type="qwen3vl_8b"):
+        super().__init__(device=device, dtype=dtype, attention_mask=attention_mask, model_options=model_options, model_type=model_type)
+        # apply the final RMSNorm to the tapped last layer
+        self.layer_norm_hidden_state = True
+
+
+class BooguTEModel(sd1_clip.SD1ClipModel):
+    def __init__(self, device="cpu", dtype=None, model_options={}):
+        clip_model = lambda **kw: BooguQwen3VLClipModel(**kw, model_type="qwen3vl_8b")
+        super().__init__(device=device, dtype=dtype, name="qwen3vl_8b", clip_model=clip_model, model_options=model_options)
+
+
+def te(dtype_llama=None, llama_quantization_metadata=None):
+    class BooguTEModel_(BooguTEModel):
+        def __init__(self, device="cpu", dtype=None, model_options={}):
+            if dtype_llama is not None:
+                dtype = dtype_llama
+            if llama_quantization_metadata is not None:
+                model_options = model_options.copy()
+                model_options["quantization_metadata"] = llama_quantization_metadata
+            super().__init__(device=device, dtype=dtype, model_options=model_options)
+    return BooguTEModel_

comfy/text_encoders/ideogram4.py

@@ -0,0 +1,120 @@
+"""Ideogram 4 text encoder: Qwen3-VL-8B language model, 13-layer tap.
+
+Ideogram 4 conditions on the concatenation of hidden states from 13 layers of
+Qwen3-VL (layers 0,3,...,33,35), giving a 4096*13 = 53248-dim feature per token.
+"""
+
+import os
+
+from transformers import Qwen2Tokenizer
+
+import comfy.text_encoders.llama
+import comfy.text_encoders.qwen3vl
+from comfy import sd1_clip
+
+# Reference taps outputs of layers (0,3,...,35); comfy captures layer inputs, offset by +1.
+IDEOGRAM4_TAP_LAYERS = [1, 4, 7, 10, 13, 16, 19, 22, 25, 28, 31, 34, 36]
+
+
+class Qwen3VLTokenizer(sd1_clip.SDTokenizer):
+    def __init__(self, embedding_directory=None, tokenizer_data={}):
+        tokenizer_path = os.path.join(os.path.dirname(os.path.realpath(__file__)), "qwen25_tokenizer")
+        super().__init__(tokenizer_path, pad_with_end=False, embedding_directory=embedding_directory,
+                         embedding_size=4096, embedding_key='qwen3vl_8b', tokenizer_class=Qwen2Tokenizer,
+                         has_start_token=False, has_end_token=False, pad_to_max_length=False,
+                         max_length=99999999, min_length=1, pad_token=151643, tokenizer_data=tokenizer_data)
+
+
+class Ideogram4Tokenizer(sd1_clip.SD1Tokenizer):
+    def __init__(self, embedding_directory=None, tokenizer_data={}):
+        super().__init__(embedding_directory=embedding_directory, tokenizer_data=tokenizer_data,
+                         name="qwen3vl_8b", tokenizer=Qwen3VLTokenizer)
+
+        self.llama_template = "<|im_start|>user\n{}<|im_end|>\n<|im_start|>assistant\n"
+
+    def tokenize_with_weights(self, text, return_word_ids=False, llama_template=None, **kwargs):
+        if text.startswith('<|im_start|>'):
+            llama_text = text
+        elif llama_template is None:
+            llama_text = self.llama_template.format(text)
+        else:
+            llama_text = llama_template.format(text)
+        return super().tokenize_with_weights(llama_text, return_word_ids=return_word_ids, disable_weights=True, **kwargs)
+
+
+# Qwen3-VL-8B = 5e6 (vs plain Qwen3-8B's 1e6)
+# final_norm/lm_head off -> Ideogram only reads raw tapped hidden states
+QWEN3VL_8B_CONFIG = {"rope_theta": 5000000.0, "final_norm": False, "lm_head": False}
+
+
+class Qwen3VL8BModel(sd1_clip.SDClipModel):
+    def __init__(self, device="cpu", layer="hidden", layer_idx=None, dtype=None, attention_mask=True, model_options={}):
+        super().__init__(device=device, layer=IDEOGRAM4_TAP_LAYERS, layer_idx=None,
+                         textmodel_json_config=dict(QWEN3VL_8B_CONFIG),
+                         dtype=dtype, special_tokens={"pad": 151643}, layer_norm_hidden_state=False,
+                         model_class=comfy.text_encoders.llama.Qwen3_8B,
+                         enable_attention_masks=attention_mask, return_attention_masks=attention_mask,
+                         model_options=model_options)
+
+
+class Ideogram4TEModel(sd1_clip.SD1ClipModel):
+    def __init__(self, device="cpu", dtype=None, model_options={}):
+        super().__init__(device=device, dtype=dtype, name="qwen3vl_8b", clip_model=Qwen3VL8BModel, model_options=model_options)
+
+    def encode_token_weights(self, token_weight_pairs):
+        out, pooled, extra = super().encode_token_weights(token_weight_pairs)
+        b, n, seq, h = out.shape # (B, n_taps=13, seq, 4096) stacked in ascending layer order.
+        out = out.permute(0, 2, 3, 1).reshape(b, seq, h * n) # (B, seq, 4096*13). permute -> (B, seq, H, taps).
+        return out, pooled, extra
+
+
+def te(dtype_llama=None, llama_quantization_metadata=None):
+    class Ideogram4TEModel_(Ideogram4TEModel):
+        def __init__(self, device="cpu", dtype=None, model_options={}):
+            if dtype_llama is not None:
+                dtype = dtype_llama
+            if llama_quantization_metadata is not None:
+                model_options = model_options.copy()
+                model_options["quantization_metadata"] = llama_quantization_metadata
+            super().__init__(device=device, dtype=dtype, model_options=model_options)
+    return Ideogram4TEModel_
+
+
+# Full Qwen3-VL-8B variant with vision
+
+class Ideogram4Qwen3VLClipModel(comfy.text_encoders.qwen3vl.Qwen3VLClipModel):
+    def __init__(self, device="cpu", dtype=None, attention_mask=True, model_options={}):
+        super().__init__(device=device, layer=IDEOGRAM4_TAP_LAYERS, layer_idx=None, dtype=dtype,
+                         attention_mask=attention_mask, model_options=model_options, model_type="qwen3vl_8b")
+
+
+class Ideogram4Qwen3VLTEModel(sd1_clip.SD1ClipModel):
+    def __init__(self, device="cpu", dtype=None, model_options={}):
+        super().__init__(device=device, dtype=dtype, name="qwen3vl_8b", clip_model=Ideogram4Qwen3VLClipModel, model_options=model_options)
+
+    def encode_token_weights(self, token_weight_pairs):
+        out, pooled, extra = super().encode_token_weights(token_weight_pairs)
+        b, n, seq, h = out.shape  # (B, n_taps=13, seq, 4096), ascending layer order.
+        out = out.permute(0, 2, 3, 1).reshape(b, seq, h * n)  # (B, seq, 4096*13 = 53248).
+        return out, pooled, extra
+
+
+class Ideogram4Qwen3VLTokenizer(comfy.text_encoders.qwen3vl.Qwen3VLTokenizer):
+    def __init__(self, embedding_directory=None, tokenizer_data={}):
+        super().__init__(embedding_directory=embedding_directory, tokenizer_data=tokenizer_data, model_type="qwen3vl_8b")
+
+    def tokenize_with_weights(self, text, return_word_ids=False, llama_template=None, images=[], prevent_empty_text=False, thinking=True, **kwargs):
+        # Ideogram 4 conditions on the no-think template; default thinking=True drops the empty think block qwen3vl adds.
+        return super().tokenize_with_weights(text, return_word_ids=return_word_ids, llama_template=llama_template, images=images, prevent_empty_text=prevent_empty_text, thinking=thinking, **kwargs)
+
+
+def te_qwen3vl(dtype_llama=None, llama_quantization_metadata=None):
+    class Ideogram4Qwen3VLTEModel_(Ideogram4Qwen3VLTEModel):
+        def __init__(self, device="cpu", dtype=None, model_options={}):
+            if dtype_llama is not None:
+                dtype = dtype_llama
+            if llama_quantization_metadata is not None:
+                model_options = model_options.copy()
+                model_options["quantization_metadata"] = llama_quantization_metadata
+            super().__init__(device=device, dtype=dtype, model_options=model_options)
+    return Ideogram4Qwen3VLTEModel_

comfy/text_encoders/llama.py

@@ -251,6 +251,19 @@ class Qwen3_8BConfig:
     lm_head: bool = True
     stop_tokens = [151643, 151645]
 
+@dataclass
+class Qwen3VL_8BConfig(Qwen3_8BConfig):
+    max_position_embeddings: int = 262144
+    rope_theta: float = 5000000.0
+    rope_dims = [24, 20, 20]
+    interleaved_mrope = True
+
+@dataclass
+class Qwen3VL_4BConfig(Qwen3VL_8BConfig):
+    hidden_size: int = 2560
+    intermediate_size: int = 9728
+    lm_head: bool = False  # 4B ties word embeddings
+
 @dataclass
 class Ovis25_2BConfig:
     vocab_size: int = 151936
@@ -703,7 +716,8 @@ class Llama2_(nn.Module):
                                     interleaved_mrope=getattr(self.config, "interleaved_mrope", False),
                                     device=device)
 
-    def forward(self, x, attention_mask=None, embeds=None, num_tokens=None, intermediate_output=None, final_layer_norm_intermediate=True, dtype=None, position_ids=None, embeds_info=[], past_key_values=None, input_ids=None):
+    def forward(self, x, attention_mask=None, embeds=None, num_tokens=None, intermediate_output=None, final_layer_norm_intermediate=True,
+                dtype=None, position_ids=None, embeds_info=[], past_key_values=None, input_ids=None,deepstack_embeds=None, visual_pos_masks=None):
         if embeds is not None:
             x = embeds
         else:
@@ -767,6 +781,10 @@ class Llama2_(nn.Module):
             if current_kv is not None:
                 next_key_values.append(current_kv)
 
+            # DeepStack: add per-layer visual features into the first len() decoder layers at image positions (Qwen3-VL)
+            if deepstack_embeds is not None and i < len(deepstack_embeds):
+                x[visual_pos_masks] = x[visual_pos_masks] + deepstack_embeds[i].to(x)
+
             if i == intermediate_output:
                 intermediate = x.clone()
 
@@ -860,7 +878,7 @@ class BaseGenerate:
                                     torch.empty([batch, model_config.num_key_value_heads, max_cache_len, model_config.head_dim], device=device, dtype=execution_dtype), 0))
         return past_key_values
 
-    def generate(self, embeds=None, do_sample=True, max_length=256, temperature=1.0, top_k=50, top_p=0.9, min_p=0.0, repetition_penalty=1.0, seed=42, stop_tokens=None, initial_tokens=[], execution_dtype=None, min_tokens=0, presence_penalty=0.0, initial_input_ids=None):
+    def generate(self, embeds=None, do_sample=True, max_length=256, temperature=1.0, top_k=50, top_p=0.9, min_p=0.0, repetition_penalty=1.0, seed=42, stop_tokens=None, initial_tokens=[], execution_dtype=None, min_tokens=0, presence_penalty=0.0, initial_input_ids=None, position_ids=None, deepstack_embeds=None, visual_pos_masks=None):
         device = embeds.device
 
         if stop_tokens is None:
@@ -884,10 +902,18 @@ class BaseGenerate:
         generated_token_ids = []
         pbar = comfy.utils.ProgressBar(max_length)
 
+        # MRoPE: prefill uses explicit 3D position_ids, decode continues from the last position
+        next_pos = int(position_ids[:, -1].max()) + 1 if position_ids is not None else None
+
         # Generation loop
         current_input_ids = initial_input_ids
         for step in tqdm(range(max_length), desc="Generating tokens"):
-            x, _, past_key_values = self.model.forward(None, embeds=embeds, attention_mask=None, past_key_values=past_key_values, input_ids=current_input_ids)
+            # DeepStack visual features are injected on the prefill only; gemma4's forward lacks these kwargs.
+            extra = {}
+            if step == 0 and deepstack_embeds is not None:
+                extra["deepstack_embeds"] = deepstack_embeds
+                extra["visual_pos_masks"] = visual_pos_masks
+            x, _, past_key_values = self.model.forward(None, embeds=embeds, attention_mask=None, past_key_values=past_key_values, input_ids=current_input_ids, position_ids=position_ids, **extra)
             logits = self.logits(x)[:, -1]
             next_token = self.sample_token(logits, temperature, top_k, top_p, min_p, repetition_penalty, initial_tokens + generated_token_ids, generator, do_sample=do_sample, presence_penalty=presence_penalty)
             token_id = next_token[0].item()
@@ -895,6 +921,9 @@ class BaseGenerate:
 
             embeds = self.model.embed_tokens(next_token).to(execution_dtype)
             current_input_ids = next_token if initial_input_ids is not None else None
+            if next_pos is not None:  # advance MRoPE position for the next (decode) step
+                position_ids = torch.tensor([[next_pos]], device=device)
+                next_pos += 1
             pbar.update(1)
 
             if token_id in stop_tokens:

comfy/text_encoders/qwen35.py

@@ -3,7 +3,6 @@ import torch.nn as nn
 import torch.nn.functional as F
 from dataclasses import dataclass, field
 import os
-import math
 
 import comfy.model_management
 from comfy.ldm.modules.attention import optimized_attention_for_device
@@ -563,6 +562,8 @@ class Qwen35VisionModel(nn.Module):
             for _ in range(config["depth"])
         ])
         self.merger = Qwen35VisionPatchMerger(self.hidden_size, self.spatial_merge_size, config["out_hidden_size"], device=device, dtype=dtype, ops=ops)
+        self.deepstack_visual_indexes = [] # DeepStack, per-layer visual features (Qwen3-VL)
+        self.deepstack_merger_list = None
 
     def rot_pos_emb(self, grid_thw):
         merge_size = self.spatial_merge_size
@@ -664,9 +665,14 @@ class Qwen35VisionModel(nn.Module):
         ).cumsum(dim=0, dtype=torch.int32)
         cu_seqlens = F.pad(cu_seqlens, (1, 0), value=0)
         optimized_attention = optimized_attention_for_device(x.device, mask=False, small_input=True)
-        for blk in self.blocks:
+        deepstack_features = []
+        for layer_num, blk in enumerate(self.blocks):
             x = blk(x, cu_seqlens=cu_seqlens, position_embeddings=position_embeddings, optimized_attention=optimized_attention)
+            if self.deepstack_merger_list is not None and layer_num in self.deepstack_visual_indexes:
+                deepstack_features.append(self.deepstack_merger_list[self.deepstack_visual_indexes.index(layer_num)](x))
         merged = self.merger(x)
+        if self.deepstack_merger_list is not None:
+            return merged, deepstack_features
         return merged
 
 # Model Wrapper
@@ -690,30 +696,7 @@ class Qwen35(BaseLlama, BaseGenerate, torch.nn.Module):
         return None, None
 
     def forward(self, x, attention_mask=None, embeds=None, num_tokens=None, intermediate_output=None, final_layer_norm_intermediate=True, dtype=None, embeds_info=[], past_key_values=None):
-        grid = None
-        position_ids = None
-        offset = 0
-        for e in embeds_info:
-            if e.get("type") == "image":
-                grid = e.get("extra", None)
-                start = e.get("index")
-                if position_ids is None:
-                    position_ids = torch.zeros((3, embeds.shape[1]), device=embeds.device)
-                    position_ids[:, :start] = torch.arange(0, start, device=embeds.device)
-                end = e.get("size") + start
-                len_max = int(grid.max()) // 2
-                start_next = len_max + start
-                position_ids[:, end:] = torch.arange(start_next + offset, start_next + (embeds.shape[1] - end) + offset, device=embeds.device)
-                position_ids[0, start:end] = start + offset
-                max_d = int(grid[0][1]) // 2
-                position_ids[1, start:end] = torch.arange(start + offset, start + max_d + offset, device=embeds.device).unsqueeze(1).repeat(1, math.ceil((end - start) / max_d)).flatten(0)[:end - start]
-                max_d = int(grid[0][2]) // 2
-                position_ids[2, start:end] = torch.arange(start + offset, start + max_d + offset, device=embeds.device).unsqueeze(0).repeat(math.ceil((end - start) / max_d), 1).flatten(0)[:end - start]
-                offset += len_max - (end - start)
-
-        if grid is None:
-            position_ids = None
-
+        position_ids = comfy.text_encoders.qwen_vl.qwen2vl_mrope_position_ids(embeds_info, embeds.shape[1], embeds.device)
         return super().forward(x, attention_mask=attention_mask, embeds=embeds, num_tokens=num_tokens, intermediate_output=intermediate_output, final_layer_norm_intermediate=final_layer_norm_intermediate, dtype=dtype, position_ids=position_ids, past_key_values=past_key_values)
 
     def init_kv_cache(self, batch, max_cache_len, device, execution_dtype):

comfy/text_encoders/qwen3vl.py

@@ -0,0 +1,193 @@
+import os
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from transformers import Qwen2Tokenizer
+
+from comfy import sd1_clip
+import comfy.text_encoders.qwen_vl
+from .qwen35 import Qwen35VisionModel
+from .llama import BaseLlama, BaseQwen3, BaseGenerate, Llama2_, Qwen3VL_4BConfig, Qwen3VL_8BConfig
+
+
+QWEN3VL_VISION = {
+    "qwen3vl_4b": dict(hidden_size=1024, intermediate_size=4096, depth=24, deepstack_visual_indexes=[5, 11, 17]),
+    "qwen3vl_8b": dict(hidden_size=1152, intermediate_size=4304, depth=27, deepstack_visual_indexes=[8, 16, 24]),
+}
+QWEN3VL_VISION_COMMON = dict(num_heads=16, patch_size=16, temporal_patch_size=2, in_channels=3,
+                             spatial_merge_size=2, num_position_embeddings=2304)
+
+QWEN3VL_CONFIGS = {"qwen3vl_4b": Qwen3VL_4BConfig, "qwen3vl_8b": Qwen3VL_8BConfig}
+
+
+class Qwen3VLDeepstackMerger(nn.Module):
+    # DeepStack merger: postshuffle LayerNorm (applied after spatial merge), unlike the main merger.
+    def __init__(self, hidden_size, spatial_merge_size, out_hidden_size, device=None, dtype=None, ops=None):
+        super().__init__()
+        self.merge_dim = hidden_size * (spatial_merge_size ** 2)
+        self.norm = ops.LayerNorm(self.merge_dim, eps=1e-6, device=device, dtype=dtype)
+        self.linear_fc1 = ops.Linear(self.merge_dim, self.merge_dim, device=device, dtype=dtype)
+        self.linear_fc2 = ops.Linear(self.merge_dim, out_hidden_size, device=device, dtype=dtype)
+
+    def forward(self, x):
+        x = self.norm(x.view(-1, self.merge_dim))
+        return self.linear_fc2(F.gelu(self.linear_fc1(x)))
+
+
+class Qwen3VLVisionModel(Qwen35VisionModel):
+    # Qwen3.5 vision + DeepStack
+    def __init__(self, config, device=None, dtype=None, ops=None):
+        super().__init__(config, device=device, dtype=dtype, ops=ops)
+        self.deepstack_visual_indexes = config["deepstack_visual_indexes"]
+        self.deepstack_merger_list = nn.ModuleList([
+            Qwen3VLDeepstackMerger(self.hidden_size, self.spatial_merge_size, config["out_hidden_size"], device=device, dtype=dtype, ops=ops)
+            for _ in self.deepstack_visual_indexes
+        ])
+
+
+class Qwen3VL(BaseLlama, BaseQwen3, BaseGenerate, torch.nn.Module):
+    model_type = "qwen3vl_8b"
+
+    def __init__(self, config_dict, dtype, device, operations):
+        super().__init__()
+        config = QWEN3VL_CONFIGS[self.model_type](**config_dict)
+        self.num_layers = config.num_hidden_layers
+        self.model = Llama2_(config, device=device, dtype=dtype, ops=operations)
+        vision_config = {**QWEN3VL_VISION_COMMON, **QWEN3VL_VISION[self.model_type], "out_hidden_size": config.hidden_size}
+        self.visual = Qwen3VLVisionModel(vision_config, device=device, dtype=dtype, ops=operations)
+        self.dtype = dtype
+
+    def preprocess_embed(self, embed, device):
+        if embed["type"] == "image":
+            # Qwen3-VL normalizes to [-1, 1] (mean/std 0.5), unlike Qwen2.5-VL's CLIP normalization.
+            image, grid = comfy.text_encoders.qwen_vl.process_qwen2vl_images(embed["data"], patch_size=16, image_mean=[0.5, 0.5, 0.5], image_std=[0.5, 0.5, 0.5])
+            merged, deepstack = self.visual(image.to(device, dtype=torch.float32), grid)
+            return merged, {"grid": grid, "deepstack": deepstack}
+        return None, None
+
+    def build_image_inputs(self, embeds, embeds_info):
+        # Returns (position_ids, visual_pos_masks, deepstack) for the prompt
+        images = sorted([e for e in embeds_info if e.get("type") == "image"], key=lambda e: e["index"])
+        if len(images) == 0:
+            return None, None, None
+
+        device = embeds.device
+        seq = embeds.shape[1]
+        position_ids = comfy.text_encoders.qwen_vl.qwen2vl_mrope_position_ids(embeds_info, seq, device)
+
+        # DeepStack: mask of image positions + per-vision-layer features to inject there.
+        visual_pos_masks = torch.zeros((1, seq), dtype=torch.bool, device=device)
+        deepstack = None
+        for e in images:
+            start = e["index"]
+            end = e["size"] + start
+            visual_pos_masks[0, start:end] = True
+            ds = e["extra"]["deepstack"]
+            if deepstack is None:
+                deepstack = [d for d in ds]
+            else:
+                deepstack = [torch.cat([deepstack[i], ds[i]], dim=0) for i in range(len(ds))]
+        return position_ids, visual_pos_masks, deepstack
+
+
+def _make_qwen3vl_model(model_type):
+    class Qwen3VL_(Qwen3VL):
+        pass
+    Qwen3VL_.model_type = model_type
+    return Qwen3VL_
+
+
+class Qwen3VLClipModel(sd1_clip.SDClipModel):
+    def __init__(self, device="cpu", layer="hidden", layer_idx=-1, dtype=None, attention_mask=True, model_options={}, model_type="qwen3vl_8b"):
+        super().__init__(device=device, layer=layer, layer_idx=layer_idx, textmodel_json_config={},
+                         dtype=dtype, special_tokens={"pad": 151643}, layer_norm_hidden_state=False,
+                         model_class=_make_qwen3vl_model(model_type), enable_attention_masks=attention_mask,
+                         return_attention_masks=attention_mask, model_options=model_options)
+
+    def generate(self, tokens, do_sample, max_length, temperature, top_k, top_p, min_p, repetition_penalty, seed, presence_penalty=0.0):
+        if isinstance(tokens, dict):
+            tokens = next(iter(tokens.values()))
+        tokens_only = [[t[0] for t in b] for b in tokens]
+        embeds, _, _, embeds_info = self.process_tokens(tokens_only, self.execution_device)
+        position_ids, visual_pos_masks, deepstack = self.transformer.build_image_inputs(embeds, embeds_info)
+        return self.transformer.generate(embeds, do_sample, max_length, temperature, top_k, top_p, min_p, repetition_penalty, seed,
+                                         presence_penalty=presence_penalty, position_ids=position_ids,
+                                         visual_pos_masks=visual_pos_masks, deepstack_embeds=deepstack)
+
+
+class Qwen3VLTEModel(sd1_clip.SD1ClipModel):
+    def __init__(self, device="cpu", dtype=None, model_options={}, model_type="qwen3vl_8b"):
+        clip_model = lambda **kw: Qwen3VLClipModel(**kw, model_type=model_type)
+        super().__init__(device=device, dtype=dtype, name=model_type, clip_model=clip_model, model_options=model_options)
+
+
+class Qwen3VLSDTokenizer(sd1_clip.SDTokenizer):
+    def __init__(self, embedding_directory=None, tokenizer_data={}, embedding_size=4096, embedding_key="qwen3vl_8b"):
+        tokenizer_path = os.path.join(os.path.dirname(os.path.realpath(__file__)), "qwen25_tokenizer")
+        super().__init__(tokenizer_path, pad_with_end=False, embedding_directory=embedding_directory, embedding_size=embedding_size, embedding_key=embedding_key, tokenizer_class=Qwen2Tokenizer,
+                         has_start_token=False, has_end_token=False, pad_to_max_length=False, max_length=99999999, min_length=1, pad_token=151643, tokenizer_data=tokenizer_data)
+
+
+class Qwen3VLTokenizer(sd1_clip.SD1Tokenizer):
+    def __init__(self, embedding_directory=None, tokenizer_data={}, model_type="qwen3vl_8b"):
+        embedding_size = 2560 if model_type == "qwen3vl_4b" else 4096
+        tokenizer = lambda *a, **kw: Qwen3VLSDTokenizer(*a, **kw, embedding_size=embedding_size, embedding_key=model_type)
+        super().__init__(embedding_directory=embedding_directory, tokenizer_data=tokenizer_data, name=model_type, tokenizer=tokenizer)
+        self.llama_template = "<|im_start|>user\n{}<|im_end|>\n<|im_start|>assistant\n"
+        self.llama_template_images = "<|im_start|>user\n<|vision_start|><|image_pad|><|vision_end|>{}<|im_end|>\n<|im_start|>assistant\n"
+
+    def tokenize_with_weights(self, text, return_word_ids=False, llama_template=None, images=[], prevent_empty_text=False, thinking=False, **kwargs):
+        image = kwargs.get("image", None)
+        if image is not None and len(images) == 0:
+            images = [image[i:i + 1] for i in range(image.shape[0])]
+
+        skip_template = text.startswith('<|im_start|>')
+        if prevent_empty_text and text == '':
+            text = ' '
+
+        if skip_template:
+            llama_text = text
+        else:
+            if llama_template is not None:
+                template = llama_template
+            elif len(images) == 0:
+                template = self.llama_template
+            else:
+                template = self.llama_template_images
+                if len(images) > 1:
+                    vision_block = "<|vision_start|><|image_pad|><|vision_end|>"
+                    template = template.replace(vision_block, vision_block * len(images), 1)
+            llama_text = template.format(text)
+            if not thinking:  # Qwen3 convention: empty think block suppresses reasoning
+                llama_text += "<think>\n\n</think>\n\n"
+
+        tokens = super().tokenize_with_weights(llama_text, return_word_ids=return_word_ids, disable_weights=True, **kwargs)
+        key_name = next(iter(tokens))
+        embed_count = 0
+        for r in tokens[key_name]:
+            for i in range(len(r)):
+                if r[i][0] == 151655:  # <|image_pad|>
+                    if len(images) > embed_count:
+                        r[i] = ({"type": "image", "data": images[embed_count], "original_type": "image"},) + r[i][1:]
+                        embed_count += 1
+        return tokens
+
+
+def tokenizer(model_type="qwen3vl_8b"):
+    class Qwen3VLTokenizer_(Qwen3VLTokenizer):
+        def __init__(self, embedding_directory=None, tokenizer_data={}):
+            super().__init__(embedding_directory=embedding_directory, tokenizer_data=tokenizer_data, model_type=model_type)
+    return Qwen3VLTokenizer_
+
+
+def te(dtype_llama=None, llama_quantization_metadata=None, model_type="qwen3vl_8b"):
+    class Qwen3VLTEModel_(Qwen3VLTEModel):
+        def __init__(self, device="cpu", dtype=None, model_options={}):
+            if dtype_llama is not None:
+                dtype = dtype_llama
+            if llama_quantization_metadata is not None:
+                model_options = model_options.copy()
+                model_options["quantization_metadata"] = llama_quantization_metadata
+            super().__init__(device=device, dtype=dtype, model_options=model_options, model_type=model_type)
+    return Qwen3VLTEModel_

comfy/text_encoders/qwen_vl.py

@@ -88,6 +88,32 @@ def process_qwen2vl_images(
     return flatten_patches, image_grid_thw
 
 
+def qwen2vl_mrope_position_ids(embeds_info, seq_len, device):
+    # (3, seq_len) T/H/W MRoPE position ids: text runs sequentially, each image span gets its grid positions.
+    # Returns None when there are no image embeds. `extra` is the image grid_thw, or a dict carrying it under "grid".
+    position_ids = None
+    offset = 0
+    for e in embeds_info:
+        if e.get("type") == "image":
+            extra = e.get("extra", None)
+            grid = extra["grid"] if isinstance(extra, dict) else extra
+            start = e.get("index")
+            if position_ids is None:
+                position_ids = torch.zeros((3, seq_len), device=device)
+                position_ids[:, :start] = torch.arange(0, start, device=device)
+            end = e.get("size") + start
+            len_max = int(grid.max()) // 2
+            start_next = len_max + start
+            position_ids[:, end:] = torch.arange(start_next + offset, start_next + (seq_len - end) + offset, device=device)
+            position_ids[0, start:end] = start + offset
+            max_d = int(grid[0][1]) // 2
+            position_ids[1, start:end] = torch.arange(start + offset, start + max_d + offset, device=device).unsqueeze(1).repeat(1, math.ceil((end - start) / max_d)).flatten(0)[:end - start]
+            max_d = int(grid[0][2]) // 2
+            position_ids[2, start:end] = torch.arange(start + offset, start + max_d + offset, device=device).unsqueeze(0).repeat(math.ceil((end - start) / max_d), 1).flatten(0)[:end - start]
+            offset += len_max - (end - start)
+    return position_ids
+
+
 class VisionPatchEmbed(nn.Module):
     def __init__(
         self,

comfy/utils.py

@@ -85,9 +85,9 @@ _TYPES = {
 def load_safetensors(ckpt):
     import comfy_aimdo.model_mmap
 
-    f = open(ckpt, "rb", buffering=0)
     file_lock = threading.Lock()
     model_mmap = comfy_aimdo.model_mmap.ModelMMAP(ckpt)
+    f = model_mmap.get_file_handle()
     file_size = os.path.getsize(ckpt)
     mv = memoryview((ctypes.c_uint8 * file_size).from_address(model_mmap.get()))
 
@@ -1452,3 +1452,10 @@ def deepcopy_list_dict(obj, memo=None):
 
     memo[obj_id] = res
     return res
+
+def bit_reverse_range(index, bits):
+    result = 0
+    for _ in range(bits):
+        result = (result << 1) | (index & 1)
+        index >>= 1
+    return result

comfy_api/feature_flags.py

@@ -25,6 +25,11 @@ CLI_FEATURE_FLAG_REGISTRY: dict[str, FeatureFlagInfo] = {
         "default": False,
         "description": "Show the sign-in button in the frontend even when not signed in",
     },
+    "enable_telemetry": {
+        "type": "bool",
+        "default": False,
+        "description": "Signal the frontend that telemetry collection is enabled",
+    },
 }
 
 

comfy_api/latest/__init__.py

@@ -5,7 +5,7 @@ from comfy_api.internal.singleton import ProxiedSingleton
 from comfy_api.internal.async_to_sync import create_sync_class
 from ._input import ImageInput, AudioInput, MaskInput, LatentInput, VideoInput
 from ._input_impl import VideoFromFile, VideoFromComponents
-from ._util import VideoCodec, VideoContainer, VideoComponents, MESH, VOXEL, File3D
+from ._util import VideoCodec, VideoContainer, VideoComponents, MESH, VOXEL, SPLAT, File3D
 from . import _io_public as io
 from . import _ui_public as ui
 from comfy_execution.utils import get_executing_context
@@ -143,6 +143,7 @@ class Types:
     VideoComponents = VideoComponents
     MESH = MESH
     VOXEL = VOXEL
+    SPLAT = SPLAT
     File3D = File3D
 
 

comfy_api/latest/_input/video_types.py

@@ -27,10 +27,13 @@ class VideoInput(ABC):
         path: Union[str, IO[bytes]],
         format: VideoContainer = VideoContainer.AUTO,
         codec: VideoCodec = VideoCodec.AUTO,
-        metadata: Optional[dict] = None
+        metadata: Optional[dict] = None,
+        bit_depth: int | None = None,
     ):
         """
         Abstract method to save the video input to a file.
+
+        bit_depth selects the encoded bit depth; None keeps the video's native depth.
         """
         pass
 
@@ -65,6 +68,12 @@ class VideoInput(ABC):
         buffer.seek(0)
         return buffer
 
+    def get_active_trim_window(self) -> tuple[float, float]:
+        """Return the active trim as ``(start_time, duration)`` in seconds (start_time normalized
+        to ``>= 0``; ``duration == 0`` means "until the end"). Default: no trim; trimmable subclasses override.
+        """
+        return 0.0, 0.0
+
     # Provide a default implementation, but subclasses can provide optimized versions
     # if possible.
     def get_dimensions(self) -> tuple[int, int]:
@@ -77,6 +86,14 @@ class VideoInput(ABC):
         components = self.get_components()
         return components.images.shape[2], components.images.shape[1]
 
+    def get_bit_depth(self) -> int:
+        """
+        Returns the bit depth of the video (e.g. 8 or 10).
+
+        Default implementation returns 8; subclasses report their real depth.
+        """
+        return 8
+
     def get_duration(self) -> float:
         """
         Returns the duration of the video in seconds.

comfy_api/latest/_input_impl/video_types.py

@@ -52,6 +52,12 @@ def get_open_write_kwargs(
     return open_kwargs
 
 
+def video_stream_bit_depth(stream) -> int:
+    if stream is None or stream.format is None or not stream.format.components:
+        return 8
+    return max(component.bits for component in stream.format.components)
+
+
 class VideoFromFile(VideoInput):
     """
     Class representing video input from a file.
@@ -75,6 +81,12 @@ class VideoFromFile(VideoInput):
             self.__file.seek(0)
         return self.__file
 
+    def get_active_trim_window(self) -> tuple[float, float]:
+        start_time = self.__start_time
+        if start_time < 0:
+            start_time = max(self._get_raw_duration() + start_time, 0.0)
+        return float(start_time), float(self.__duration)
+
     def get_dimensions(self) -> tuple[int, int]:
         """
         Returns the dimensions of the video input.
@@ -91,6 +103,13 @@ class VideoFromFile(VideoInput):
                     return stream.width, stream.height
         raise ValueError(f"No video stream found in file '{self.__file}'")
 
+    def get_bit_depth(self) -> int:
+        if isinstance(self.__file, io.BytesIO):
+            self.__file.seek(0)  # Reset the BytesIO object to the beginning
+        with av.open(self.__file, mode="r") as container:
+            video_stream = container.streams.video[0] if len(container.streams.video) > 0 else None
+            return video_stream_bit_depth(video_stream)
+
     def get_duration(self) -> float:
         """
         Returns the duration of the video in seconds.
@@ -251,6 +270,7 @@ class VideoFromFile(VideoInput):
 
         image_format = 'gbrpf32le'
         process_image_format = lambda a: a
+        align_graph = None
         audio = None
 
         streams = [video_stream]
@@ -304,7 +324,28 @@ class VideoFromFile(VideoInput):
 
                             checked_alpha = True
 
-                        img = frame.to_ndarray(format=image_format)  # shape: (H, W, 4)
+                        # Fix non-deterministic video decode when the video width is not a multiple of 32
+                        # For non-yuvj pixel formats: most H.264/H.265 video and static images (e.g. lossy WebP via LoadImage)
+                        # Pad both axes to a multiple of 32 and smear the border so the alignment padding never bleeds into the cropped edges
+                        if image_format in ('gbrpf32le', 'gbrapf32le') and frame.width % 32 != 0:
+                            if align_graph is None:
+                                pad_w = ((frame.width + 31) // 32) * 32
+                                pad_h = ((frame.height + 31) // 32) * 32
+                                g = av.filter.Graph()
+                                g_src = g.add_buffer(width=frame.width, height=frame.height,
+                                                     format=frame.format.name, time_base=video_stream.time_base)
+                                g_pad = g.add('pad', f'{pad_w}:{pad_h}:0:0')
+                                g_fill = g.add('fillborders', f'left=0:right={pad_w - frame.width}:top=0:bottom={pad_h - frame.height}:mode=smear')
+                                g_sink = g.add('buffersink')
+                                g_src.link_to(g_pad)
+                                g_pad.link_to(g_fill)
+                                g_fill.link_to(g_sink)
+                                g.configure()
+                                align_graph = (g, g_src, g_sink)
+                            align_graph[1].push(frame)
+                            img = np.ascontiguousarray(align_graph[2].pull().to_ndarray(format=image_format)[:frame.height, :frame.width])
+                        else:
+                            img = frame.to_ndarray(format=image_format)
                         if frame.rotation != 0:
                             k = int(round(frame.rotation // 90))
                             img = np.rot90(img, k=k, axes=(0, 1)).copy()
@@ -371,25 +412,32 @@ class VideoFromFile(VideoInput):
         format: VideoContainer = VideoContainer.AUTO,
         codec: VideoCodec = VideoCodec.AUTO,
         metadata: Optional[dict] = None,
+        bit_depth: int | None = None,
     ):
         if isinstance(self.__file, io.BytesIO):
             self.__file.seek(0)  # Reset the BytesIO object to the beginning
         with av.open(self.__file, mode='r') as container:
             container_format = container.format.name
-            video_encoding = container.streams.video[0].codec.name if len(container.streams.video) > 0 else None
+            video_stream = container.streams.video[0] if len(container.streams.video) > 0 else None
+            video_encoding = video_stream.codec.name if video_stream is not None else None
+            source_bit_depth = video_stream_bit_depth(video_stream)
             reuse_streams = True
             if format != VideoContainer.AUTO and format not in container_format.split(","):
                 reuse_streams = False
             if codec != VideoCodec.AUTO and codec != video_encoding and video_encoding is not None:
                 reuse_streams = False
+            if bit_depth is not None and video_encoding is not None and bit_depth != source_bit_depth:
+                reuse_streams = False
             if self.__start_time or self.__duration:
                 reuse_streams = False
 
             if not reuse_streams:
+                if bit_depth is None:
+                    bit_depth = source_bit_depth
                 components = self.get_components_internal(container)
                 video = VideoFromComponents(components)
                 return video.save_to(
-                    path, format=format, codec=codec, metadata=metadata
+                    path, format=format, codec=codec, metadata=metadata, bit_depth=bit_depth,
                 )
 
             streams = container.streams
@@ -445,8 +493,10 @@ class VideoFromComponents(VideoInput):
     Class representing video input from tensors.
     """
 
-    def __init__(self, components: VideoComponents):
+    def __init__(self, components: VideoComponents, bit_depth: int = 8):
         self.__components = components
+        # Tensor components have no inherent bit depth; this is the depth used when encoding.
+        self.__bit_depth = bit_depth
 
     def get_components(self) -> VideoComponents:
         return VideoComponents(
@@ -455,18 +505,26 @@ class VideoFromComponents(VideoInput):
             frame_rate=self.__components.frame_rate,
         )
 
+    def get_bit_depth(self) -> int:
+        return self.__bit_depth
+
     def save_to(
         self,
         path: str,
         format: VideoContainer = VideoContainer.AUTO,
         codec: VideoCodec = VideoCodec.AUTO,
         metadata: Optional[dict] = None,
+        bit_depth: int | None = None,
     ):
         """Save the video to a file path or BytesIO buffer."""
         if format != VideoContainer.AUTO and format != VideoContainer.MP4:
             raise ValueError("Only MP4 format is supported for now")
         if codec != VideoCodec.AUTO and codec != VideoCodec.H264:
             raise ValueError("Only H264 codec is supported for now")
+        # None means "use the depth this video was created with" (CreateVideo's choice).
+        if bit_depth is None:
+            bit_depth = self.__bit_depth
+        is_10bit = bit_depth >= 10
         extra_kwargs = {}
         if isinstance(format, VideoContainer) and format != VideoContainer.AUTO:
             extra_kwargs["format"] = format.value
@@ -482,10 +540,11 @@ class VideoFromComponents(VideoInput):
 
             frame_rate = Fraction(round(self.__components.frame_rate * 1000), 1000)
             # Create a video stream
+            pix_fmt = "yuv420p10le" if is_10bit else "yuv420p"
             video_stream = output.add_stream('h264', rate=frame_rate)
             video_stream.width = self.__components.images.shape[2]
             video_stream.height = self.__components.images.shape[1]
-            video_stream.pix_fmt = 'yuv420p'
+            video_stream.pix_fmt = pix_fmt
 
             # Create an audio stream
             audio_sample_rate = 1
@@ -499,9 +558,14 @@ class VideoFromComponents(VideoInput):
 
             # Encode video
             for i, frame in enumerate(self.__components.images):
-                img = (frame * 255).clamp(0, 255).byte().cpu().numpy() # shape: (H, W, 3)
-                frame = av.VideoFrame.from_ndarray(img, format='rgb24')
-                frame = frame.reformat(format='yuv420p')  # Convert to YUV420P as required by h264
+                if is_10bit:
+                    # 16-bit RGB keeps float precision through the conversion to 10-bit YUV.
+                    img = (frame.float() * 65535).clamp(0, 65535).cpu().numpy().astype(np.uint16)  # shape: (H, W, 3)
+                    frame = av.VideoFrame.from_ndarray(img, format="rgb48le")
+                else:
+                    img = (frame * 255).clamp(0, 255).byte().cpu().numpy() # shape: (H, W, 3)
+                    frame = av.VideoFrame.from_ndarray(img, format='rgb24')
+                frame = frame.reformat(format=pix_fmt)
                 packet = video_stream.encode(frame)
                 output.mux(packet)
 

comfy_api/latest/_io.py

@@ -28,7 +28,7 @@ if TYPE_CHECKING:
 from comfy_api.internal import (_ComfyNodeInternal, _NodeOutputInternal, classproperty, copy_class, first_real_override, is_class,
     prune_dict, shallow_clone_class)
 from comfy_execution.graph_utils import ExecutionBlocker
-from ._util import MESH, VOXEL, SVG as _SVG, File3D
+from ._util import MESH, VOXEL, SPLAT, SVG as _SVG, File3D
 
 
 class FolderType(str, Enum):
@@ -684,6 +684,10 @@ class Voxel(ComfyTypeIO):
 class Mesh(ComfyTypeIO):
     Type = MESH
 
+@comfytype(io_type="SPLAT")
+class Splat(ComfyTypeIO):
+    Type = SPLAT
+
 
 @comfytype(io_type="FILE_3D")
 class File3DAny(ComfyTypeIO):
@@ -727,6 +731,42 @@ class File3DUSDZ(ComfyTypeIO):
     Type = File3D
 
 
+@comfytype(io_type="FILE_3D_PLY")
+class File3DPLY(ComfyTypeIO):
+    """PLY format 3D file - point cloud or Gaussian splat."""
+    Type = File3D
+
+
+@comfytype(io_type="FILE_3D_SPLAT")
+class File3DSPLAT(ComfyTypeIO):
+    """SPLAT format 3D file - 3D Gaussian splat."""
+    Type = File3D
+
+
+@comfytype(io_type="FILE_3D_SPZ")
+class File3DSPZ(ComfyTypeIO):
+    """SPZ format 3D file - compressed 3D Gaussian splat."""
+    Type = File3D
+
+
+@comfytype(io_type="FILE_3D_KSPLAT")
+class File3DKSPLAT(ComfyTypeIO):
+    """KSPLAT format 3D file - 3D Gaussian splat."""
+    Type = File3D
+
+
+@comfytype(io_type="FILE_3D_SPLAT_ANY")
+class File3DSplatAny(ComfyTypeIO):
+    """General 3D Gaussian splat file type - accepts any supported splat container (.ply / .spz / .splat / .ksplat)."""
+    Type = File3D
+
+
+@comfytype(io_type="FILE_3D_POINT_CLOUD_ANY")
+class File3DPointCloudAny(ComfyTypeIO):
+    """General point cloud file type - accepts any supported point cloud container (currently .ply)."""
+    Type = File3D
+
+
 @comfytype(io_type="HOOKS")
 class Hooks(ComfyTypeIO):
     if TYPE_CHECKING:
@@ -1360,7 +1400,8 @@ class V3Data(TypedDict):
 class HiddenHolder:
     def __init__(self, unique_id: str, prompt: Any,
                  extra_pnginfo: Any, dynprompt: Any,
-                 auth_token_comfy_org: str, api_key_comfy_org: str, **kwargs):
+                 auth_token_comfy_org: str, api_key_comfy_org: str,
+                 comfy_usage_source: str = None, **kwargs):
         self.unique_id = unique_id
         """UNIQUE_ID is the unique identifier of the node, and matches the id property of the node on the client side. It is commonly used in client-server communications (see messages)."""
         self.prompt = prompt
@@ -1373,6 +1414,8 @@ class HiddenHolder:
         """AUTH_TOKEN_COMFY_ORG is a token acquired from signing into a ComfyOrg account on frontend."""
         self.api_key_comfy_org = api_key_comfy_org
         """API_KEY_COMFY_ORG is an API Key generated by ComfyOrg that allows skipping signing into a ComfyOrg account on frontend."""
+        self.comfy_usage_source = comfy_usage_source
+        """COMFY_USAGE_SOURCE identifies the client that submitted the prompt (e.g. comfyui-frontend, comfy-cli, comfyui-mcp); forwarded to API nodes' upstream requests via the Comfy-Usage-Source header."""
 
     def __getattr__(self, key: str):
         '''If hidden variable not found, return None.'''
@@ -1389,6 +1432,7 @@ class HiddenHolder:
             dynprompt=d.get(Hidden.dynprompt, None),
             auth_token_comfy_org=d.get(Hidden.auth_token_comfy_org, None),
             api_key_comfy_org=d.get(Hidden.api_key_comfy_org, None),
+            comfy_usage_source=d.get(Hidden.comfy_usage_source, None),
         )
 
     @classmethod
@@ -1411,6 +1455,8 @@ class Hidden(str, Enum):
     """AUTH_TOKEN_COMFY_ORG is a token acquired from signing into a ComfyOrg account on frontend."""
     api_key_comfy_org = "API_KEY_COMFY_ORG"
     """API_KEY_COMFY_ORG is an API Key generated by ComfyOrg that allows skipping signing into a ComfyOrg account on frontend."""
+    comfy_usage_source = "COMFY_USAGE_SOURCE"
+    """COMFY_USAGE_SOURCE identifies the client that submitted the prompt (e.g. comfyui-frontend, comfy-cli, comfyui-mcp); forwarded to API nodes' upstream requests via the Comfy-Usage-Source header."""
 
 
 @dataclass
@@ -1614,6 +1660,8 @@ class Schema:
                 self.hidden.append(Hidden.auth_token_comfy_org)
             if Hidden.api_key_comfy_org not in self.hidden:
                 self.hidden.append(Hidden.api_key_comfy_org)
+            if Hidden.comfy_usage_source not in self.hidden:
+                self.hidden.append(Hidden.comfy_usage_source)
         # if is an output_node, will need prompt and extra_pnginfo
         if self.is_output_node:
             if Hidden.prompt not in self.hidden:
@@ -2296,6 +2344,7 @@ __all__ = [
     "LossMap",
     "Voxel",
     "Mesh",
+    "Splat",
     "File3DAny",
     "File3DGLB",
     "File3DGLTF",
@@ -2303,6 +2352,12 @@ __all__ = [
     "File3DOBJ",
     "File3DSTL",
     "File3DUSDZ",
+    "File3DPLY",
+    "File3DSPLAT",
+    "File3DSPZ",
+    "File3DKSPLAT",
+    "File3DSplatAny",
+    "File3DPointCloudAny",
     "Hooks",
     "HookKeyframes",
     "TimestepsRange",

comfy_api/latest/_ui.py

@@ -285,7 +285,7 @@ class AudioSaveHelper:
         results = []
         for batch_number, waveform in enumerate(audio["waveform"].cpu()):
             filename_with_batch_num = filename.replace("%batch_num%", str(batch_number))
-            file = f"{filename_with_batch_num}_{counter:05}_.{format}"
+            file = f"{filename_with_batch_num}_{counter:05}.{format}"
             output_path = os.path.join(full_output_folder, file)
 
             # Use original sample rate initially
@@ -452,6 +452,16 @@ class PreviewUI3D(_UIOutput):
         return {"result": [self.model_file, self.camera_info, self.bg_image_path]}
 
 
+class PreviewUI3DAdvanced(_UIOutput):
+    def __init__(self, model_file, camera_info, model_3d_info):
+        self.model_file = model_file
+        self.camera_info = camera_info
+        self.model_3d_info = model_3d_info
+
+    def as_dict(self):
+        return {"result": [self.model_file, self.camera_info, self.model_3d_info]}
+
+
 class PreviewText(_UIOutput):
     def __init__(self, value: str, **kwargs):
         self.value = value
@@ -471,5 +481,6 @@ __all__ = [
     "PreviewAudio",
     "PreviewVideo",
     "PreviewUI3D",
+    "PreviewUI3DAdvanced",
     "PreviewText",
 ]

comfy_api/latest/_util/__init__.py

@@ -1,5 +1,5 @@
 from .video_types import VideoContainer, VideoCodec, VideoComponents
-from .geometry_types import VOXEL, MESH, File3D
+from .geometry_types import VOXEL, MESH, SPLAT, File3D
 from .image_types import SVG
 
 __all__ = [
@@ -9,6 +9,7 @@ __all__ = [
     "VideoComponents",
     "VOXEL",
     "MESH",
+    "SPLAT",
     "File3D",
     "SVG",
 ]

comfy_api/latest/_util/geometry_types.py

@@ -11,13 +11,32 @@ class VOXEL:
         self.data = data
 
 
+class SPLAT:
+    """A batch of 3D Gaussian splats in render-ready (activated, world-space) form.
+
+    Tensors are (B, N, ...) and zero-padded to a common N across the batch; `counts` (B,) holds the
+    real per-item lengths (None when rows are uniform and no slicing is needed). SH coefficients are
+    stored as (B, N, K, 3) with K = (sh_degree + 1)**2; the DC (diffuse) term is sh[..., 0, :].
+    """
+
+    def __init__(self, positions: torch.Tensor, scales: torch.Tensor, rotations: torch.Tensor,
+                 opacities: torch.Tensor, sh: torch.Tensor, counts: torch.Tensor | None = None):
+        self.positions = positions    # (B, N, 3) world-space centers
+        self.scales = scales          # (B, N, 3) linear (positive) per-axis std
+        self.rotations = rotations    # (B, N, 4) quaternion wxyz (normalized)
+        self.opacities = opacities    # (B, N, 1) in [0, 1]
+        self.sh = sh                  # (B, N, K, 3) spherical-harmonic color coefficients
+        self.counts = counts          # (B,) real lengths, or None
+
+
 class MESH:
     def __init__(self, vertices: torch.Tensor, faces: torch.Tensor,
                  uvs: torch.Tensor | None = None,
                  vertex_colors: torch.Tensor | None = None,
                  texture: torch.Tensor | None = None,
                  vertex_counts: torch.Tensor | None = None,
-                 face_counts: torch.Tensor | None = None):
+                 face_counts: torch.Tensor | None = None,
+                 unlit: bool = False):
 
         assert (vertex_counts is None) == (face_counts is None), \
             "vertex_counts and face_counts must be provided together (both or neither)"
@@ -30,6 +49,8 @@ class MESH:
         # these hold the real per-item lengths (B,). None means rows are uniform and no slicing is needed.
         self.vertex_counts = vertex_counts
         self.face_counts = face_counts
+        # Render flat / emissive (no scene lighting) when saved, e.g. for gaussian-splat-derived meshes.
+        self.unlit = unlit
 
 
 class File3D:

comfy_api_nodes/apis/__init__.py

@@ -1310,13 +1310,6 @@ class KlingTaskStatus(str, Enum):
     failed = 'failed'
 
 
-class KlingTextToVideoModelName(str, Enum):
-    kling_v1 = 'kling-v1'
-    kling_v1_6 = 'kling-v1-6'
-    kling_v2_1_master = 'kling-v2-1-master'
-    kling_v2_5_turbo = 'kling-v2-5-turbo'
-
-
 class KlingVideoGenAspectRatio(str, Enum):
     field_16_9 = '16:9'
     field_9_16 = '9:16'
@@ -5179,7 +5172,7 @@ class KlingText2VideoRequest(BaseModel):
     duration: Optional[KlingVideoGenDuration] = '5'
     external_task_id: Optional[str] = Field(None, description='Customized Task ID')
     mode: Optional[KlingVideoGenMode] = 'std'
-    model_name: Optional[KlingTextToVideoModelName] = 'kling-v1'
+    model_name: Optional[str] = 'kling-v1'
     negative_prompt: Optional[str] = Field(
         None, description='Negative text prompt', max_length=2500
     )

comfy_api_nodes/apis/bfl.py

@@ -1,71 +1,72 @@
 from enum import Enum
-from typing import Any, Dict, Optional
+from typing import Any
 
-from pydantic import BaseModel, Field, confloat, conint
-
-
-class BFLOutputFormat(str, Enum):
-    png = 'png'
-    jpeg = 'jpeg'
+from pydantic import BaseModel, Field
 
 
 class BFLFluxExpandImageRequest(BaseModel):
-    prompt: str = Field(..., description='The description of the changes you want to make. This text guides the expansion process, allowing you to specify features, styles, or modifications for the expanded areas.')
-    prompt_upsampling: Optional[bool] = Field(
-        None, description='Whether to perform upsampling on the prompt. If active, automatically modifies the prompt for more creative generation.'
-    )
-    seed: Optional[int] = Field(None, description='The seed value for reproducibility.')
-    top: conint(ge=0, le=2048) = Field(..., description='Number of pixels to expand at the top of the image')
-    bottom: conint(ge=0, le=2048) = Field(..., description='Number of pixels to expand at the bottom of the image')
-    left: conint(ge=0, le=2048) = Field(..., description='Number of pixels to expand at the left side of the image')
-    right: conint(ge=0, le=2048) = Field(..., description='Number of pixels to expand at the right side of the image')
-    steps: conint(ge=15, le=50) = Field(..., description='Number of steps for the image generation process')
-    guidance: confloat(ge=1.5, le=100) = Field(..., description='Guidance strength for the image generation process')
-    safety_tolerance: Optional[conint(ge=0, le=6)] = Field(
-        6, description='Tolerance level for input and output moderation. Between 0 and 6, 0 being most strict, 6 being least strict. Defaults to 2.'
-    )
-    output_format: Optional[BFLOutputFormat] = Field(
-        BFLOutputFormat.png, description="Output format for the generated image. Can be 'jpeg' or 'png'.", examples=['png']
-    )
-    image: str = Field(None, description='A Base64-encoded string representing the image you wish to expand')
+    prompt: str = Field(...)
+    prompt_upsampling: bool | None = Field(None)
+    seed: int | None = Field(None)
+    top: int = Field(...)
+    bottom: int = Field(...)
+    left: int = Field(...)
+    right: int = Field(...)
+    steps: int = Field(...)
+    guidance: float = Field(...)
+    safety_tolerance: int = Field(6)
+    output_format: str = Field("png")
+    image: str = Field(None, description="A Base64-encoded string representing the image you wish to expand")
 
 
 class BFLFluxFillImageRequest(BaseModel):
-    prompt: str = Field(..., description='The description of the changes you want to make. This text guides the expansion process, allowing you to specify features, styles, or modifications for the expanded areas.')
-    prompt_upsampling: Optional[bool] = Field(
-        None, description='Whether to perform upsampling on the prompt. If active, automatically modifies the prompt for more creative generation.'
+    prompt: str = Field(...)
+    prompt_upsampling: bool | None = Field(None)
+    seed: int | None = Field(None)
+    steps: int = Field(...)
+    guidance: float = Field(...)
+    safety_tolerance: int = Field(6)
+    output_format: str = Field("png")
+    image: str = Field(
+        None, description="Base64-encoded string representing the image to modify. Can contain alpha mask if desired.",
     )
-    seed: Optional[int] = Field(None, description='The seed value for reproducibility.')
-    steps: conint(ge=15, le=50) = Field(..., description='Number of steps for the image generation process')
-    guidance: confloat(ge=1.5, le=100) = Field(..., description='Guidance strength for the image generation process')
-    safety_tolerance: Optional[conint(ge=0, le=6)] = Field(
-        6, description='Tolerance level for input and output moderation. Between 0 and 6, 0 being most strict, 6 being least strict. Defaults to 2.'
+    mask: str = Field(
+        None, description="Base64-encoded string representing the mask of the areas you wish to modify."
     )
-    output_format: Optional[BFLOutputFormat] = Field(
-        BFLOutputFormat.png, description="Output format for the generated image. Can be 'jpeg' or 'png'.", examples=['png']
+
+
+class BFLFluxEraseRequest(BaseModel):
+    image: str = Field(..., description="A Base64-encoded string representing the image to erase from.")
+    mask: str = Field(
+        ...,
+        description="A Base64-encoded black/white mask matching the input dimensions; "
+        "white (255) marks areas to remove, black (0) marks areas to preserve.",
     )
-    image: str = Field(None, description='A Base64-encoded string representing the image you wish to modify. Can contain alpha mask if desired.')
-    mask: str = Field(None, description='A Base64-encoded string representing the mask of the areas you with to modify.')
+    dilate_pixels: int = Field(10)
+    seed: int | None = Field(None)
+    output_format: str = Field("png")
+
+
+class BFLFluxVTORequest(BaseModel):
+    prompt: str = Field(
+        ..., description="Natural-language styling instruction. Required field, but may be an empty string."
+    )
+    person: str = Field(..., description="A Base64-encoded string representing the person image.")
+    garment: str = Field(..., description="A Base64-encoded string representing the garment reference image.")
+    seed: int | None = Field(None)
+    safety_tolerance: int = Field(5)
+    output_format: str = Field("png")
 
 
 class BFLFluxProGenerateRequest(BaseModel):
-    prompt: str = Field(..., description='The text prompt for image generation.')
-    prompt_upsampling: Optional[bool] = Field(
-        None, description='Whether to perform upsampling on the prompt. If active, automatically modifies the prompt for more creative generation.'
-    )
-    seed: Optional[int] = Field(None, description='The seed value for reproducibility.')
-    width: conint(ge=256, le=1440) = Field(1024, description='Width of the generated image in pixels. Must be a multiple of 32.')
-    height: conint(ge=256, le=1440) = Field(768, description='Height of the generated image in pixels. Must be a multiple of 32.')
-    safety_tolerance: Optional[conint(ge=0, le=6)] = Field(
-        6, description='Tolerance level for input and output moderation. Between 0 and 6, 0 being most strict, 6 being least strict. Defaults to 2.'
-    )
-    output_format: Optional[BFLOutputFormat] = Field(
-        BFLOutputFormat.png, description="Output format for the generated image. Can be 'jpeg' or 'png'.", examples=['png']
-    )
-    image_prompt: Optional[str] = Field(None, description='Optional image to remix in base64 format')
-    # image_prompt_strength: Optional[confloat(ge=0.0, le=1.0)] = Field(
-    #     None, description='Blend between the prompt and the image prompt.'
-    # )
+    prompt: str = Field(...)
+    prompt_upsampling: bool | None = Field(None)
+    seed: int | None = Field(None)
+    width: int = Field(1024, description="Must be a multiple of 32.")
+    height: int = Field(768, description="Must be a multiple of 32.")
+    safety_tolerance: int = Field(6)
+    output_format: str = Field("png")
+    image_prompt: str | None = Field(None, description="Optional image to remix in base64 format")
 
 
 class Flux2ProGenerateRequest(BaseModel):
@@ -83,55 +84,37 @@ class Flux2ProGenerateRequest(BaseModel):
     input_image_7: str | None = Field(None, description="Base64 encoded image for image-to-image generation")
     input_image_8: str | None = Field(None, description="Base64 encoded image for image-to-image generation")
     input_image_9: str | None = Field(None, description="Base64 encoded image for image-to-image generation")
-    safety_tolerance: int | None = Field(
-        5, description="Tolerance level for input and output moderation. Value 0 being most strict.", ge=0, le=5
-    )
-    output_format: str | None = Field(
-        "png", description="Output format for the generated image. Can be 'jpeg' or 'png'."
-    )
+    safety_tolerance: int = Field(5)
+    output_format: str = Field("png")
 
 
 class BFLFluxKontextProGenerateRequest(BaseModel):
-    prompt: str = Field(..., description='The text prompt for what you wannt to edit.')
-    input_image: Optional[str] = Field(None, description='Image to edit in base64 format')
-    seed: Optional[int] = Field(None, description='The seed value for reproducibility.')
-    guidance: confloat(ge=0.1, le=99.0) = Field(..., description='Guidance strength for the image generation process')
-    steps: conint(ge=1, le=150) = Field(..., description='Number of steps for the image generation process')
-    safety_tolerance: Optional[conint(ge=0, le=2)] = Field(
-        2, description='Tolerance level for input and output moderation. Between 0 and 2, 0 being most strict, 6 being least strict. Defaults to 2.'
-    )
-    output_format: Optional[BFLOutputFormat] = Field(
-        BFLOutputFormat.png, description="Output format for the generated image. Can be 'jpeg' or 'png'.", examples=['png']
-    )
-    aspect_ratio: Optional[str] = Field(None, description='Aspect ratio of the image between 21:9 and 9:21.')
-    prompt_upsampling: Optional[bool] = Field(
-        None, description='Whether to perform upsampling on the prompt. If active, automatically modifies the prompt for more creative generation.'
-    )
+    prompt: str = Field(...)
+    input_image: str | None = Field(None, description="Image to edit in base64 format")
+    seed: int | None = Field(None)
+    guidance: float = Field(...)
+    steps: int = Field(...)
+    safety_tolerance: int = Field(2)
+    output_format: str = Field("png")
+    aspect_ratio: str | None = Field(None)
+    prompt_upsampling: bool | None = Field(None)
 
 
 class BFLFluxProUltraGenerateRequest(BaseModel):
-    prompt: str = Field(..., description='The text prompt for image generation.')
-    prompt_upsampling: Optional[bool] = Field(
-        None, description='Whether to perform upsampling on the prompt. If active, automatically modifies the prompt for more creative generation.'
-    )
-    seed: Optional[int] = Field(None, description='The seed value for reproducibility.')
-    aspect_ratio: Optional[str] = Field(None, description='Aspect ratio of the image between 21:9 and 9:21.')
-    safety_tolerance: Optional[conint(ge=0, le=6)] = Field(
-        6, description='Tolerance level for input and output moderation. Between 0 and 6, 0 being most strict, 6 being least strict. Defaults to 2.'
-    )
-    output_format: Optional[BFLOutputFormat] = Field(
-        BFLOutputFormat.png, description="Output format for the generated image. Can be 'jpeg' or 'png'.", examples=['png']
-    )
-    raw: Optional[bool] = Field(None, description='Generate less processed, more natural-looking images.')
-    image_prompt: Optional[str] = Field(None, description='Optional image to remix in base64 format')
-    image_prompt_strength: Optional[confloat(ge=0.0, le=1.0)] = Field(
-        None, description='Blend between the prompt and the image prompt.'
-    )
+    prompt: str = Field(...)
+    prompt_upsampling: bool | None = Field(None)
+    seed: int | None = Field(None)
+    aspect_ratio: str | None = Field(None)
+    safety_tolerance: int = Field(6)
+    output_format: str = Field("png")
+    raw: bool | None = Field(None)
+    image_prompt: str | None = Field(None, description="Optional image to remix in base64 format")
+    image_prompt_strength: float | None = Field(None)
 
 
 class BFLFluxProGenerateResponse(BaseModel):
-    id: str = Field(..., description="The unique identifier for the generation task.")
-    polling_url: str = Field(..., description="URL to poll for the generation result.")
+    id: str = Field(...)
+    polling_url: str = Field(...)
     cost: float | None = Field(None, description="Price in cents")
 
 
@@ -145,7 +128,7 @@ class BFLStatus(str, Enum):
 
 
 class BFLFluxStatusResponse(BaseModel):
-    id: str = Field(..., description="The unique identifier for the generation task.")
-    status: BFLStatus = Field(..., description="The status of the task.")
-    result: Optional[Dict[str, Any]] = Field(None, description="The result of the task (null if not completed).")
-    progress: Optional[float] = Field(None, description="The progress of the task (0.0 to 1.0).", ge=0.0, le=1.0)
+    id: str = Field(...)
+    status: BFLStatus = Field(...)
+    result: dict[str, Any] | None = Field(None)
+    progress: float | None = Field(None, ge=0.0, le=1.0)

comfy_api_nodes/apis/bria.py

@@ -97,3 +97,28 @@ class BriaRemoveVideoBackgroundResult(BaseModel):
 class BriaRemoveVideoBackgroundResponse(BaseModel):
     status: str = Field(...)
     result: BriaRemoveVideoBackgroundResult | None = Field(None)
+
+
+class BriaVideoGreenScreenRequest(BaseModel):
+    video: str = Field(..., description="Publicly accessible URL of the input video.")
+    green_shade: str = Field(
+        default="broadcast_green",
+        description="Solid chroma-key shade applied behind the foreground "
+        "(broadcast_green, chroma_green, or blue_screen).",
+    )
+    output_container_and_codec: str = Field(...)
+    preserve_audio: bool = Field(True)
+    seed: int = Field(...)
+
+
+class BriaVideoReplaceBackgroundRequest(BaseModel):
+    video: str = Field(..., description="Publicly accessible URL of the input (foreground) video.")
+    background_url: str = Field(
+        ...,
+        description="Publicly accessible URL of the background image or video to composite behind "
+        "the foreground. Stretched to the foreground frame; match its aspect ratio for "
+        "undistorted results.",
+    )
+    output_container_and_codec: str = Field(...)
+    preserve_audio: bool = Field(True)
+    seed: int = Field(...)

comfy_api_nodes/apis/gemini.py

@@ -108,13 +108,19 @@ class GeminiVideoMetadata(BaseModel):
     startOffset: GeminiOffset | None = Field(None)
 
 
+class GeminiThinkingConfig(BaseModel):
+    includeThoughts: bool | None = Field(None)
+    thinkingLevel: str = Field(...)
+
+
 class GeminiGenerationConfig(BaseModel):
-    maxOutputTokens: int | None = Field(None, ge=16, le=8192)
+    maxOutputTokens: int | None = Field(None, ge=16, le=65536)
     seed: int | None = Field(None)
     stopSequences: list[str] | None = Field(None)
     temperature: float | None = Field(None, ge=0.0, le=2.0)
     topK: int | None = Field(None, ge=1)
     topP: float | None = Field(None, ge=0.0, le=1.0)
+    thinkingConfig: GeminiThinkingConfig | None = Field(None)
 
 
 class GeminiImageOutputOptions(BaseModel):
@@ -128,11 +134,6 @@ class GeminiImageConfig(BaseModel):
     imageOutputOptions: GeminiImageOutputOptions = Field(default_factory=GeminiImageOutputOptions)
 
 
-class GeminiThinkingConfig(BaseModel):
-    includeThoughts: bool | None = Field(None)
-    thinkingLevel: str = Field(...)
-
-
 class GeminiImageGenerationConfig(GeminiGenerationConfig):
     responseModalities: list[str] | None = Field(None)
     imageConfig: GeminiImageConfig | None = Field(None)

comfy_api_nodes/apis/ideogram.py

@@ -290,3 +290,19 @@ class IdeogramV3Request(BaseModel):
         None,
         description='Optional masks for character reference images. When provided, must match the number of character_reference_images. Each mask should be a grayscale image of the same dimensions as the corresponding character reference image. The images should be in JPEG, PNG or WebP format.'
     )
+
+
+class IdeogramV4Request(BaseModel):
+    text_prompt: str | None = Field(
+        None,
+        description="Natural-language prompt; Magic Prompt is applied automatically. "
+        "Supply exactly one of text_prompt or json_prompt.",
+    )
+    json_prompt: dict[str, Any] | None = Field(
+        None,
+        description="Structured V4 prompt object consumed directly (disables Magic Prompt). "
+        "Supply exactly one of text_prompt or json_prompt.",
+    )
+    resolution: str | None = Field(None, description="Output resolution in WIDTHxHEIGHT (e.g. '2048x2048').")
+    rendering_speed: str | None = Field(None, description="Rendering speed: 'TURBO', 'DEFAULT', or 'QUALITY'.")
+    enable_copyright_detection: bool | None = Field(None, description="Opt into post-generation copyright detection.")

comfy_api_nodes/apis/kling.py

@@ -149,3 +149,59 @@ class MotionControlRequest(BaseModel):
     character_orientation: str = Field(...)
     mode: str = Field(..., description="'pro' or 'std'")
     model_name: str = Field(...)
+
+
+class Kling3TurboSettings(BaseModel):
+    resolution: str = Field("720p", description="'720p' or '1080p'")
+    aspect_ratio: str | None = Field(None, description="'16:9'/'9:16'/'1:1'; text-to-video only")
+    duration: int = Field(5, description="3-15 second")
+
+
+class Kling3TurboText2VideoRequest(BaseModel):
+    prompt: str = Field(..., description="<=3072 chars; may use multi-shot 'shot n, m, words; ...'")
+    settings: Kling3TurboSettings | None = Field(None)
+
+
+class Kling3TurboContent(BaseModel):
+    type: str = Field(..., description="'prompt' or 'first_frame'")
+    text: str | None = Field(None, description="for type=prompt; <=2500 chars")
+    url: str | None = Field(None, description="for type=first_frame")
+
+
+class Kling3TurboImage2VideoRequest(BaseModel):
+    contents: list[Kling3TurboContent] = Field(..., description="prompt + first_frame materials")
+    settings: Kling3TurboSettings | None = Field(None)
+
+
+class Kling3TurboCreateData(BaseModel):
+    id: str | None = Field(None, description="Task ID")
+    status: str | None = Field(None)
+    message: str | None = Field(None)
+
+
+class Kling3TurboCreateResponse(BaseModel):
+    code: int | None = Field(None)
+    message: str | None = Field(None)
+    request_id: str | None = Field(None)
+    data: Kling3TurboCreateData | None = Field(None)
+
+
+class Kling3TurboOutput(BaseModel):
+    type: str | None = Field(None, description="'video', 'image', 'audio', ...")
+    id: str | None = Field(None)
+    url: str | None = Field(None)
+    duration: str | None = Field(None)
+
+
+class Kling3TurboTaskData(BaseModel):
+    id: str | None = Field(None)
+    status: str | None = Field(None, description="submitted | processing | succeeded | failed")
+    message: str | None = Field(None)
+    outputs: list[Kling3TurboOutput] | None = Field(None)
+
+
+class Kling3TurboQueryResponse(BaseModel):
+    code: int | None = Field(None)
+    message: str | None = Field(None)
+    request_id: str | None = Field(None)
+    data: list[Kling3TurboTaskData] | None = Field(None)

comfy_api_nodes/apis/luma.py

@@ -10,6 +10,7 @@ from pydantic import BaseModel, Field, confloat
 class LumaIO:
     LUMA_REF = "LUMA_REF"
     LUMA_CONCEPTS = "LUMA_CONCEPTS"
+    LUMA_RAY32_KEYFRAME = "LUMA_RAY32_KEYFRAME"
 
 
 class LumaReference:
@@ -20,13 +21,14 @@ class LumaReference:
     def create_api_model(self, download_url: str):
         return LumaImageRef(url=download_url, weight=self.weight)
 
+
 class LumaReferenceChain:
-    def __init__(self, first_ref: LumaReference=None):
+    def __init__(self, first_ref: LumaReference = None):
         self.refs: list[LumaReference] = []
         if first_ref:
             self.refs.append(first_ref)
 
-    def add(self, luma_ref: LumaReference=None):
+    def add(self, luma_ref: LumaReference = None):
         self.refs.append(luma_ref)
 
     def create_api_model(self, download_urls: list[str], max_refs=4):
@@ -124,7 +126,7 @@ def get_luma_concepts(include_none=False):
         "pull_out",
         "aerial",
         "crane_up",
-        "eye_level"
+        "eye_level",
     ]
 
 
@@ -162,8 +164,8 @@ class LumaVideoModelOutputDuration(str, Enum):
 
 
 class LumaGenerationType(str, Enum):
-    video = 'video'
-    image = 'image'
+    video = "video"
+    image = "image"
 
 
 class LumaState(str, Enum):
@@ -174,86 +176,109 @@ class LumaState(str, Enum):
 
 
 class LumaAssets(BaseModel):
-    video: Optional[str] = Field(None, description='The URL of the video')
-    image: Optional[str] = Field(None, description='The URL of the image')
-    progress_video: Optional[str] = Field(None, description='The URL of the progress video')
+    video: Optional[str] = Field(None, description="The URL of the video")
+    image: Optional[str] = Field(None, description="The URL of the image")
+    progress_video: Optional[str] = Field(None, description="The URL of the progress video")
 
 
 class LumaImageRef(BaseModel):
     """Used for image gen"""
-    url: str = Field(..., description='The URL of the image reference')
-    weight: confloat(ge=0.0, le=1.0) = Field(..., description='The weight of the image reference')
+
+    url: str = Field(..., description="The URL of the image reference")
+    weight: confloat(ge=0.0, le=1.0) = Field(..., description="The weight of the image reference")
 
 
 class LumaImageReference(BaseModel):
     """Used for video gen"""
-    type: Optional[str] = Field('image', description='Input type, defaults to image')
-    url: str = Field(..., description='The URL of the image')
+
+    type: Optional[str] = Field("image", description="Input type, defaults to image")
+    url: str = Field(..., description="The URL of the image")
 
 
 class LumaModifyImageRef(BaseModel):
-    url: str = Field(..., description='The URL of the image reference')
-    weight: confloat(ge=0.0, le=1.0) = Field(..., description='The weight of the image reference')
+    url: str = Field(..., description="The URL of the image reference")
+    weight: confloat(ge=0.0, le=1.0) = Field(..., description="The weight of the image reference")
 
 
 class LumaCharacterRef(BaseModel):
-    identity0: LumaImageIdentity = Field(..., description='The image identity object')
+    identity0: LumaImageIdentity = Field(..., description="The image identity object")
 
 
 class LumaImageIdentity(BaseModel):
-    images: list[str] = Field(..., description='The URLs of the image identity')
+    images: list[str] = Field(..., description="The URLs of the image identity")
 
 
 class LumaGenerationReference(BaseModel):
-    type: str = Field('generation', description='Input type, defaults to generation')
-    id: str = Field(..., description='The ID of the generation')
+    type: str = Field("generation", description="Input type, defaults to generation")
+    id: str = Field(..., description="The ID of the generation")
 
 
 class LumaKeyframes(BaseModel):
-    frame0: Optional[Union[LumaImageReference, LumaGenerationReference]] = Field(None, description='')
-    frame1: Optional[Union[LumaImageReference, LumaGenerationReference]] = Field(None, description='')
+    frame0: Optional[Union[LumaImageReference, LumaGenerationReference]] = Field(None, description="")
+    frame1: Optional[Union[LumaImageReference, LumaGenerationReference]] = Field(None, description="")
 
 
 class LumaConceptObject(BaseModel):
-    key: str = Field(..., description='Camera Concept name')
+    key: str = Field(..., description="Camera Concept name")
 
 
 class LumaImageGenerationRequest(BaseModel):
-    prompt: str = Field(..., description='The prompt of the generation')
-    model: LumaImageModel = Field(LumaImageModel.photon_1, description='The image model used for the generation')
-    aspect_ratio: Optional[LumaAspectRatio] = Field(LumaAspectRatio.ratio_16_9, description='The aspect ratio of the generation')
-    image_ref: Optional[list[LumaImageRef]] = Field(None, description='List of image reference objects')
-    style_ref: Optional[list[LumaImageRef]] = Field(None, description='List of style reference objects')
-    character_ref: Optional[LumaCharacterRef] = Field(None, description='The image identity object')
-    modify_image_ref: Optional[LumaModifyImageRef] = Field(None, description='The modify image reference object')
+    prompt: str = Field(..., description="The prompt of the generation")
+    model: LumaImageModel = Field(LumaImageModel.photon_1, description="The image model used for the generation")
+    aspect_ratio: Optional[LumaAspectRatio] = Field(LumaAspectRatio.ratio_16_9)
+    image_ref: Optional[list[LumaImageRef]] = Field(None, description="List of image reference objects")
+    style_ref: Optional[list[LumaImageRef]] = Field(None, description="List of style reference objects")
+    character_ref: Optional[LumaCharacterRef] = Field(None, description="The image identity object")
+    modify_image_ref: Optional[LumaModifyImageRef] = Field(None, description="The modify image reference object")
 
 
 class LumaGenerationRequest(BaseModel):
-    prompt: str = Field(..., description='The prompt of the generation')
-    model: LumaVideoModel = Field(LumaVideoModel.ray_2, description='The video model used for the generation')
-    duration: Optional[LumaVideoModelOutputDuration] = Field(None, description='The duration of the generation')
-    aspect_ratio: Optional[LumaAspectRatio] = Field(None, description='The aspect ratio of the generation')
-    resolution: Optional[LumaVideoOutputResolution] = Field(None, description='The resolution of the generation')
-    loop: Optional[bool] = Field(None, description='Whether to loop the video')
-    keyframes: Optional[LumaKeyframes] = Field(None, description='The keyframes of the generation')
-    concepts: Optional[list[LumaConceptObject]] = Field(None, description='Camera Concepts to apply to generation')
+    prompt: str = Field(..., description="The prompt of the generation")
+    model: LumaVideoModel = Field(LumaVideoModel.ray_2, description="The video model used for the generation")
+    duration: Optional[LumaVideoModelOutputDuration] = Field(None, description="The duration of the generation")
+    aspect_ratio: Optional[LumaAspectRatio] = Field(None, description="The aspect ratio of the generation")
+    resolution: Optional[LumaVideoOutputResolution] = Field(None, description="The resolution of the generation")
+    loop: Optional[bool] = Field(None, description="Whether to loop the video")
+    keyframes: Optional[LumaKeyframes] = Field(None, description="The keyframes of the generation")
+    concepts: Optional[list[LumaConceptObject]] = Field(None, description="Camera Concepts to apply to generation")
 
 
 class LumaGeneration(BaseModel):
-    id: str = Field(..., description='The ID of the generation')
-    generation_type: LumaGenerationType = Field(..., description='Generation type, image or video')
-    state: LumaState = Field(..., description='The state of the generation')
-    failure_reason: Optional[str] = Field(None, description='The reason for the state of the generation')
-    created_at: str = Field(..., description='The date and time when the generation was created')
-    assets: Optional[LumaAssets] = Field(None, description='The assets of the generation')
-    model: str = Field(..., description='The model used for the generation')
-    request: Union[LumaGenerationRequest, LumaImageGenerationRequest] = Field(..., description="The request used for the generation")
+    id: str = Field(..., description="The ID of the generation")
+    generation_type: LumaGenerationType = Field(..., description="Generation type, image or video")
+    state: LumaState = Field(..., description="The state of the generation")
+    failure_reason: Optional[str] = Field(None, description="The reason for the state of the generation")
+    created_at: str = Field(..., description="The date and time when the generation was created")
+    assets: Optional[LumaAssets] = Field(None, description="The assets of the generation")
+    model: str = Field(..., description="The model used for the generation")
+    request: Union[LumaGenerationRequest, LumaImageGenerationRequest] = Field(...)
 
 
 class Luma2ImageRef(BaseModel):
     url: str | None = None
     data: str | None = None
     media_type: str | None = None
+    generation_id: str | None = Field(None, description="reference a prior generation (extend / source reuse)")
+
+
+class Luma2VideoEdit(BaseModel):
+    """Edit controls for Ray 3.2 ``video_edit`` generations."""
+
+    auto_controls: bool | None = Field(None, description="derive a conditioning schedule from the source (recommended)")
+    strength: str | None = Field(None, description="'adhere_1' .. 'reimagine_3'; constrained by IO.Combo")
+
+
+class Luma2VideoOptions(BaseModel):
+    """Ray 3.2 ``video`` output settings (text / image / keyframe / edit / extend)."""
+
+    resolution: str | None = Field(None, description="360p | 540p | 720p | 1080p")
+    duration: str | None = Field(None, description="5s | 10s")
+    loop: bool | None = Field(None)
+    start_frame: Luma2ImageRef | None = Field(None)
+    end_frame: Luma2ImageRef | None = Field(None)
+    keyframes: list[Luma2ImageRef] | None = Field(None)
+    keyframe_indexes: list[int] | None = Field(None)
+    edit: Luma2VideoEdit | None = Field(None)
 
 
 class Luma2GenerationRequest(BaseModel):
@@ -266,6 +291,7 @@ class Luma2GenerationRequest(BaseModel):
     web_search: bool | None = None
     image_ref: list[Luma2ImageRef] | None = None
     source: Luma2ImageRef | None = None
+    video: Luma2VideoOptions | None = Field(None)
 
 
 class Luma2Generation(BaseModel):
@@ -277,3 +303,31 @@ class Luma2Generation(BaseModel):
     output: list[LumaImageReference] | None = None
     failure_reason: str | None = None
     failure_code: str | None = None
+
+
+# --- Ray 3.2 multi-keyframe chain ---
+
+LUMA_KEYFRAME_MODE_FRACTION = "fraction"  # value in [0.0, 1.0] of the output video duration
+LUMA_KEYFRAME_MODE_SECONDS = "seconds"  # absolute time, in seconds, from the start of the output
+
+
+class LumaRay32KeyframeItem:
+    """One guide image anchored at a position on the Ray 3.2 output timeline."""
+
+    def __init__(self, image: torch.Tensor, mode: str, value: float):
+        self.image = image
+        self.mode = mode  # LUMA_KEYFRAME_MODE_FRACTION | LUMA_KEYFRAME_MODE_SECONDS
+        self.value = value
+
+
+class LumaRay32KeyframeChain:
+    def __init__(self):
+        self.items: list[LumaRay32KeyframeItem] = []
+
+    def add(self, item: LumaRay32KeyframeItem) -> None:
+        self.items.append(item)
+
+    def clone(self) -> "LumaRay32KeyframeChain":
+        c = LumaRay32KeyframeChain()
+        c.items = list(self.items)
+        return c

comfy_api_nodes/apis/runway.py

@@ -67,15 +67,6 @@ class RunwayImageToVideoResponse(BaseModel):
     id: Optional[str] = Field(None, description='Task ID')
 
 
-class RunwayTaskStatusEnum(str, Enum):
-    SUCCEEDED = 'SUCCEEDED'
-    RUNNING = 'RUNNING'
-    FAILED = 'FAILED'
-    PENDING = 'PENDING'
-    CANCELLED = 'CANCELLED'
-    THROTTLED = 'THROTTLED'
-
-
 class RunwayTaskStatusResponse(BaseModel):
     createdAt: datetime = Field(..., description='Task creation timestamp')
     id: str = Field(..., description='Task ID')
@@ -86,7 +77,7 @@ class RunwayTaskStatusResponse(BaseModel):
         ge=0.0,
         le=1.0,
     )
-    status: RunwayTaskStatusEnum
+    status: str = Field(..., description="SUCCEEDED, RUNNING, FAILED, PENDING, CANCELLED or THROTTLED")
 
 
 class Model4(str, Enum):
@@ -125,3 +116,144 @@ class RunwayTextToImageRequest(BaseModel):
 
 class RunwayTextToImageResponse(BaseModel):
     id: Optional[str] = Field(None, description='Task ID')
+
+
+class RunwayAleph2IO:
+    """Custom socket types for chaining Aleph2 guidance images."""
+
+    KEYFRAME = "RUNWAY_ALEPH2_KEYFRAME"
+    PROMPT_IMAGE = "RUNWAY_ALEPH2_PROMPT_IMAGE"
+
+
+# Keyframe timing modes (anchored to the INPUT video). Stored on the chain item and used to
+# choose the request model below. The values match the Aleph2 keyframe union field names.
+KEYFRAME_MODE_SECONDS = "seconds"  # absolute time, in seconds, from the start of the input video
+KEYFRAME_MODE_AT = "at"  # fraction [0.0, 1.0] of the input video duration
+
+# Prompt-image position modes (anchored to the OUTPUT video). Values match the Aleph2 position `type`.
+PROMPT_IMAGE_MODE_TIMESTAMP = "timestamp"  # absolute time, in seconds, from the start of the output video
+PROMPT_IMAGE_MODE_POSITION = "position"  # fraction [0.0, 1.0] of the output video duration
+
+
+class RunwayAleph2KeyframeItem:
+    """A guidance image anchored to a point of the INPUT video (one Aleph2 ``keyframe``)."""
+
+    def __init__(self, image, mode: str, value: float):
+        self.image = image
+        self.mode = mode  # KEYFRAME_MODE_SECONDS | KEYFRAME_MODE_AT
+        self.value = value
+
+
+class RunwayAleph2KeyframeChain:
+    """An ordered collection of keyframes, built by chaining Runway Aleph2 Keyframe nodes."""
+
+    def __init__(self):
+        self.items: list[RunwayAleph2KeyframeItem] = []
+
+    def add(self, item: RunwayAleph2KeyframeItem) -> None:
+        self.items.append(item)
+
+    def clone(self) -> "RunwayAleph2KeyframeChain":
+        c = RunwayAleph2KeyframeChain()
+        c.items = list(self.items)
+        return c
+
+
+class RunwayAleph2PromptImageItem:
+    """A guidance image anchored to a point of the OUTPUT video (one Aleph2 ``promptImage``)."""
+
+    def __init__(self, image, mode: str, value: float):
+        self.image = image
+        self.mode = mode  # PROMPT_IMAGE_MODE_TIMESTAMP | PROMPT_IMAGE_MODE_POSITION
+        self.value = value
+
+
+class RunwayAleph2PromptImageChain:
+    """An ordered collection of prompt images, built by chaining Runway Aleph2 Prompt Image nodes."""
+
+    def __init__(self):
+        self.items: list[RunwayAleph2PromptImageItem] = []
+
+    def add(self, item: RunwayAleph2PromptImageItem) -> None:
+        self.items.append(item)
+
+    def clone(self) -> "RunwayAleph2PromptImageChain":
+        c = RunwayAleph2PromptImageChain()
+        c.items = list(self.items)
+        return c
+
+
+class RunwayAleph2KeyframeSeconds(BaseModel):
+    seconds: float = Field(
+        ...,
+        description="Absolute timestamp in seconds from the start of the input video when this guidance image should apply.",
+        ge=0.0,
+    )
+    uri: str = Field(...)
+
+
+class RunwayAleph2KeyframeAt(BaseModel):
+    at: float = Field(
+        ...,
+        description="Position as a fraction [0.0, 1.0] of the input video duration.",
+        ge=0.0,
+        le=1.0,
+    )
+    uri: str = Field(...)
+
+
+class RunwayAleph2TimestampPosition(BaseModel):
+    type: str = Field(default="timestamp")
+    timestampSeconds: float = Field(
+        ...,
+        description="Absolute timestamp in seconds from the start of the output video.",
+        ge=0.0,
+    )
+
+
+class RunwayAleph2RelativePosition(BaseModel):
+    type: str = Field(default="position")
+    positionPercentage: float = Field(
+        ...,
+        description="Position as a fraction [0.0, 1.0] of the total output video duration.",
+        ge=0.0,
+        le=1.0,
+    )
+
+
+class RunwayAleph2PromptImage(BaseModel):
+    position: RunwayAleph2TimestampPosition | RunwayAleph2RelativePosition
+    uri: str = Field(...)
+
+
+class RunwayAleph2ContentModeration(BaseModel):
+    publicFigureThreshold: str = Field(
+        ...,
+        description='When set to "low", the content moderation system is less strict about '
+        'recognizable public figures. One of "auto" or "low".',
+    )
+
+
+class RunwayAleph2Request(BaseModel):
+    model: str = Field(default="aleph2")
+    promptText: str = Field(
+        ...,
+        description="A non-empty string describing what should appear in the output.",
+        min_length=1,
+        max_length=1000,
+    )
+    videoUri: str = Field(...)
+    seed: int = Field(..., description="Random seed for generation", ge=0, le=4294967295)
+    contentModeration: RunwayAleph2ContentModeration = Field(...)
+    keyframes: list[RunwayAleph2KeyframeSeconds | RunwayAleph2KeyframeAt] | None = Field(
+        None,
+        description="Timed guidance images placed at specific points in the input video. Up to 5.",
+    )
+    promptImage: list[RunwayAleph2PromptImage] | None = Field(
+        None,
+        description="Up to 5 image keyframes for guiding the edit at specific points in the output video.",
+    )
+
+
+class RunwayAleph2Response(BaseModel):
+    id: str | None = Field(None, description="Task ID")

comfy_api_nodes/apis/tripo.py

@@ -208,6 +208,10 @@ class TripoMultiviewToModelRequest(BaseModel):
     quad: bool | None = Field(False, description="Whether to apply quad to the generated model")
 
 
+class TripoTexturePrompt(BaseModel):
+    text: str | None = Field(None, description="Text guidance for texture generation")
+
+
 class TripoTextureModelRequest(BaseModel):
     type: TripoTaskType = Field(TripoTaskType.TEXTURE_MODEL, description="Type of task")
     original_model_task_id: str = Field(..., description="The task ID of the original model")
@@ -219,6 +223,11 @@ class TripoTextureModelRequest(BaseModel):
     texture_alignment: TripoTextureAlignment | None = Field(
         TripoTextureAlignment.ORIGINAL_IMAGE, description="The texture alignment method"
     )
+    texture_prompt: TripoTexturePrompt | None = Field(
+        None,
+        description="Optional guidance for texturing. Required in practice for imported models, "
+        "which carry no source image to infer texture from.",
+    )
 
 
 class TripoRefineModelRequest(BaseModel):
@@ -307,6 +316,17 @@ class TripoP1MultiviewToModelRequest(TripoP1CommonRequest):
     orientation: str | None = None
 
 
+class TripoImportModelRequest(BaseModel):
+    """Request for the comfy-api composite import endpoint (/proxy/tripo/v2/openapi/import).
+
+    The model file is uploaded to ComfyUI API storage first; the backend downloads it from
+    `url`, re-uploads it to Tripo's storage and creates the import_model task server-side.
+    """
+
+    url: str = Field(..., description="ComfyUI API storage download URL of the model file")
+    format: str = Field(..., description='File format: "glb", "fbx", "obj" or "stl"')
+
+
 class TripoTaskOutput(BaseModel):
     model: str | None = Field(None, description="URL to the model")
     base_model: str | None = Field(None, description="URL to the base model")

comfy_api_nodes/nodes_anthropic.py

@@ -155,7 +155,7 @@ class ClaudeNode(IO.ComfyNode):
         return IO.Schema(
             node_id="ClaudeNode",
             display_name="Anthropic Claude",
-            category="text/partner/Anthropic",
+            category="partner/text/Anthropic",
             essentials_category="Text Generation",
             description="Generate text responses with Anthropic's Claude models. "
             "Provide a text prompt and optionally one or more images for multimodal context.",

comfy_api_nodes/nodes_beeble.py

@@ -206,7 +206,7 @@ class BeebleSwitchXVideoEdit(IO.ComfyNode):
         return IO.Schema(
             node_id="BeebleSwitchXVideoEdit",
             display_name="Beeble SwitchX Video Edit",
-            category="video/partner/Beeble",
+            category="partner/video/Beeble",
             description=(
                 "Edit a video with Beeble SwitchX. Switches anything in the scene (background, "
                 "lighting, costume) while preserving the original subject's pixels and motion. "
@@ -302,7 +302,7 @@ class BeebleSwitchXImageEdit(IO.ComfyNode):
         return IO.Schema(
             node_id="BeebleSwitchXImageEdit",
             display_name="Beeble SwitchX Image Edit",
-            category="image/partner/Beeble",
+            category="partner/image/Beeble",
             description=(
                 "Edit a single image with Beeble SwitchX. Switches anything in the scene "
                 "(background, lighting, costume) while preserving the original subject's pixels. "

comfy_api_nodes/nodes_bfl.py

@@ -4,17 +4,20 @@ from typing_extensions import override
 
 from comfy_api.latest import IO, ComfyExtension, Input
 from comfy_api_nodes.apis.bfl import (
+    BFLFluxEraseRequest,
     BFLFluxExpandImageRequest,
     BFLFluxFillImageRequest,
     BFLFluxKontextProGenerateRequest,
     BFLFluxProGenerateResponse,
     BFLFluxProUltraGenerateRequest,
     BFLFluxStatusResponse,
+    BFLFluxVTORequest,
     BFLStatus,
     Flux2ProGenerateRequest,
 )
 from comfy_api_nodes.util import (
     ApiEndpoint,
+    convert_mask_to_image,
     download_url_to_image_tensor,
     get_number_of_images,
     poll_op,
@@ -22,19 +25,11 @@ from comfy_api_nodes.util import (
     sync_op,
     tensor_to_base64_string,
     validate_aspect_ratio_string,
+    validate_image_dimensions,
     validate_string,
 )
 
 
-def convert_mask_to_image(mask: Input.Image):
-    """
-    Make mask have the expected amount of dims (4) and channels (3) to be recognized as an image.
-    """
-    mask = mask.unsqueeze(-1)
-    mask = torch.cat([mask] * 3, dim=-1)
-    return mask
-
-
 class FluxProUltraImageNode(IO.ComfyNode):
 
     @classmethod
@@ -42,7 +37,7 @@ class FluxProUltraImageNode(IO.ComfyNode):
         return IO.Schema(
             node_id="FluxProUltraImageNode",
             display_name="Flux 1.1 [pro] Ultra Image",
-            category="image/partner/BFL",
+            category="partner/image/BFL",
             description="Generates images using Flux Pro 1.1 Ultra via api based on prompt and resolution.",
             inputs=[
                 IO.String.Input(
@@ -160,7 +155,7 @@ class FluxKontextProImageNode(IO.ComfyNode):
         return IO.Schema(
             node_id=cls.NODE_ID,
             display_name=cls.DISPLAY_NAME,
-            category="image/partner/BFL",
+            category="partner/image/BFL",
             description="Edits images using Flux.1 Kontext [pro] via api based on prompt and aspect ratio.",
             inputs=[
                 IO.String.Input(
@@ -282,7 +277,7 @@ class FluxProExpandNode(IO.ComfyNode):
         return IO.Schema(
             node_id="FluxProExpandNode",
             display_name="Flux.1 Expand Image",
-            category="image/partner/BFL",
+            category="partner/image/BFL",
             description="Outpaints image based on prompt.",
             inputs=[
                 IO.Image.Input("image"),
@@ -419,7 +414,7 @@ class FluxProFillNode(IO.ComfyNode):
         return IO.Schema(
             node_id="FluxProFillNode",
             display_name="Flux.1 Fill Image",
-            category="image/partner/BFL",
+            category="partner/image/BFL",
             description="Inpaints image based on mask and prompt.",
             inputs=[
                 IO.Image.Input("image"),
@@ -519,6 +514,174 @@ class FluxProFillNode(IO.ComfyNode):
         return IO.NodeOutput(await download_url_to_image_tensor(response.result["sample"]))
 
 
+class FluxEraseNode(IO.ComfyNode):
+
+    @classmethod
+    def define_schema(cls) -> IO.Schema:
+        return IO.Schema(
+            node_id="FluxEraseNode",
+            display_name="Flux Erase Image",
+            category="partner/image/BFL",
+            description="Removes the masked object from an image and reconstructs the background. "
+            "Paint the mask over what you want to erase.",
+            inputs=[
+                IO.Image.Input("image"),
+                IO.Mask.Input("mask", tooltip="White areas are removed; black areas are preserved."),
+                IO.Int.Input(
+                    "dilate_pixels",
+                    default=10,
+                    min=0,
+                    max=25,
+                    tooltip="Expands the mask boundaries to ensure clean coverage of the object's edges.",
+                ),
+                IO.Int.Input(
+                    "seed",
+                    default=0,
+                    min=0,
+                    max=2147483647,
+                    control_after_generate=True,
+                    tooltip="The random seed used for creating the noise.",
+                    optional=True,
+                ),
+            ],
+            outputs=[IO.Image.Output()],
+            hidden=[
+                IO.Hidden.auth_token_comfy_org,
+                IO.Hidden.api_key_comfy_org,
+                IO.Hidden.unique_id,
+            ],
+            is_api_node=True,
+            price_badge=IO.PriceBadge(
+                expr="""{"type":"range_usd","min_usd":0.03,"max_usd":0.06,"format":{"approximate":true}}""",
+            ),
+        )
+
+    @classmethod
+    async def execute(
+        cls,
+        image: Input.Image,
+        mask: Input.Image,
+        dilate_pixels: int = 10,
+        seed: int = 0,
+    ) -> IO.NodeOutput:
+        validate_image_dimensions(image, min_width=256, min_height=256)
+        mask = resize_mask_to_image(mask, image)
+        mask = tensor_to_base64_string(convert_mask_to_image(mask))
+        initial_response = await sync_op(
+            cls,
+            ApiEndpoint(path="/proxy/bfl/v1/flux-tools/erase-v1", method="POST"),
+            response_model=BFLFluxProGenerateResponse,
+            data=BFLFluxEraseRequest(
+                image=tensor_to_base64_string(image[:, :, :, :3]),  # make sure image will have alpha channel removed
+                mask=mask,
+                dilate_pixels=dilate_pixels,
+                seed=seed,
+            ),
+        )
+
+        def price_extractor(_r: BaseModel) -> float | None:
+            return None if initial_response.cost is None else initial_response.cost / 100
+
+        response = await poll_op(
+            cls,
+            ApiEndpoint(initial_response.polling_url),
+            response_model=BFLFluxStatusResponse,
+            status_extractor=lambda r: r.status,
+            progress_extractor=lambda r: r.progress,
+            price_extractor=price_extractor,
+            completed_statuses=[BFLStatus.ready],
+            failed_statuses=[
+                BFLStatus.request_moderated,
+                BFLStatus.content_moderated,
+                BFLStatus.error,
+                BFLStatus.task_not_found,
+            ],
+            queued_statuses=[],
+        )
+        return IO.NodeOutput(await download_url_to_image_tensor(response.result["sample"]))
+
+
+class FluxVTONode(IO.ComfyNode):
+
+    @classmethod
+    def define_schema(cls) -> IO.Schema:
+        return IO.Schema(
+            node_id="FluxVTONode",
+            display_name="Flux Virtual Try-On",
+            category="partner/image/BFL",
+            description="Virtual try-on: dresses the person in the provided garment.",
+            inputs=[
+                IO.Image.Input("person", tooltip="Image of the person to dress."),
+                IO.Image.Input("garment", tooltip="Image of the garment to apply."),
+                IO.String.Input(
+                    "prompt",
+                    multiline=True,
+                    default="",
+                    tooltip="Optional natural-language styling instruction (e.g. how the garment should fit).",
+                ),
+                IO.Int.Input(
+                    "seed",
+                    default=0,
+                    min=0,
+                    max=0xFFFFFFFFFFFFFFFF,
+                    control_after_generate=True,
+                    tooltip="The random seed used for creating the noise.",
+                ),
+            ],
+            outputs=[IO.Image.Output()],
+            hidden=[
+                IO.Hidden.auth_token_comfy_org,
+                IO.Hidden.api_key_comfy_org,
+                IO.Hidden.unique_id,
+            ],
+            is_api_node=True,
+            price_badge=IO.PriceBadge(
+                expr="""{"type":"range_usd","min_usd":0.0375,"max_usd":0.075,"format":{"approximate":true}}""",
+            ),
+        )
+
+    @classmethod
+    async def execute(
+        cls,
+        person: Input.Image,
+        garment: Input.Image,
+        prompt: str = "",
+        seed: int = 0,
+    ) -> IO.NodeOutput:
+        initial_response = await sync_op(
+            cls,
+            ApiEndpoint(path="/proxy/bfl/v1/flux-tools/vto-v1", method="POST"),
+            response_model=BFLFluxProGenerateResponse,
+            data=BFLFluxVTORequest(
+                prompt=prompt,
+                person=tensor_to_base64_string(person[:, :, :, :3]),
+                garment=tensor_to_base64_string(garment[:, :, :, :3]),
+                seed=seed,
+            ),
+        )
+
+        def price_extractor(_r: BaseModel) -> float | None:
+            return None if initial_response.cost is None else initial_response.cost / 100
+
+        response = await poll_op(
+            cls,
+            ApiEndpoint(initial_response.polling_url),
+            response_model=BFLFluxStatusResponse,
+            status_extractor=lambda r: r.status,
+            progress_extractor=lambda r: r.progress,
+            price_extractor=price_extractor,
+            completed_statuses=[BFLStatus.ready],
+            failed_statuses=[
+                BFLStatus.request_moderated,
+                BFLStatus.content_moderated,
+                BFLStatus.error,
+                BFLStatus.task_not_found,
+            ],
+            queued_statuses=[],
+        )
+        return IO.NodeOutput(await download_url_to_image_tensor(response.result["sample"]))
+
+
 class Flux2ProImageNode(IO.ComfyNode):
 
     NODE_ID = "Flux2ProImageNode"
@@ -545,7 +708,7 @@ class Flux2ProImageNode(IO.ComfyNode):
         return IO.Schema(
             node_id=cls.NODE_ID,
             display_name=cls.DISPLAY_NAME,
-            category="image/partner/BFL",
+            category="partner/image/BFL",
             description="Generates images synchronously based on prompt and resolution.",
             inputs=[
                 IO.String.Input(
@@ -716,7 +879,7 @@ class Flux2ImageNode(IO.ComfyNode):
         return IO.Schema(
             node_id="Flux2ImageNode",
             display_name="Flux.2 Image",
-            category="image/partner/BFL",
+            category="partner/image/BFL",
             description="Generate images via Flux.2 [pro] or Flux.2 [max] from a prompt and optional reference images.",
             inputs=[
                 IO.String.Input(
@@ -853,6 +1016,8 @@ class BFLExtension(ComfyExtension):
             FluxKontextMaxImageNode,
             FluxProExpandNode,
             FluxProFillNode,
+            FluxEraseNode,
+            FluxVTONode,
             Flux2ProImageNode,
             Flux2MaxImageNode,
             Flux2ImageNode,

comfy_api_nodes/nodes_bria.py

@@ -1,14 +1,19 @@
+import av
+import torch
+from av.codec import CodecContext
 from typing_extensions import override
 
 from comfy_api.latest import IO, ComfyExtension, Input
 from comfy_api_nodes.apis.bria import (
     BriaEditImageRequest,
+    BriaImageEditResponse,
     BriaRemoveBackgroundRequest,
     BriaRemoveBackgroundResponse,
     BriaRemoveVideoBackgroundRequest,
     BriaRemoveVideoBackgroundResponse,
-    BriaImageEditResponse,
     BriaStatusResponse,
+    BriaVideoGreenScreenRequest,
+    BriaVideoReplaceBackgroundRequest,
     InputModerationSettings,
 )
 from comfy_api_nodes.util import (
@@ -31,7 +36,7 @@ class BriaImageEditNode(IO.ComfyNode):
         return IO.Schema(
             node_id="BriaImageEditNode",
             display_name="Bria FIBO Image Edit",
-            category="image/partner/Bria",
+            category="partner/image/Bria",
             description="Edit images using Bria latest model",
             inputs=[
                 IO.Combo.Input("model", options=["FIBO"]),
@@ -169,7 +174,7 @@ class BriaRemoveImageBackground(IO.ComfyNode):
         return IO.Schema(
             node_id="BriaRemoveImageBackground",
             display_name="Bria Remove Image Background",
-            category="image/partner/Bria",
+            category="partner/image/Bria",
             description="Remove the background from an image using Bria RMBG 2.0.",
             inputs=[
                 IO.Image.Input("image"),
@@ -245,7 +250,7 @@ class BriaRemoveVideoBackground(IO.ComfyNode):
         return IO.Schema(
             node_id="BriaRemoveVideoBackground",
             display_name="Bria Remove Video Background",
-            category="video/partner/Bria",
+            category="partner/video/Bria",
             description="Remove the background from a video using Bria. ",
             inputs=[
                 IO.Video.Input("video"),
@@ -284,7 +289,7 @@ class BriaRemoveVideoBackground(IO.ComfyNode):
             ],
             is_api_node=True,
             price_badge=IO.PriceBadge(
-                expr="""{"type":"usd","usd":0.14,"format":{"suffix":"/second"}}""",
+                expr="""{"type":"usd","usd":0.0042,"format":{"suffix":"/second"}}""",
             ),
         )
 
@@ -316,6 +321,251 @@ class BriaRemoveVideoBackground(IO.ComfyNode):
         return IO.NodeOutput(await download_url_to_video_output(response.result.video_url))
 
 
+class BriaVideoGreenScreen(IO.ComfyNode):
+
+    @classmethod
+    def define_schema(cls):
+        return IO.Schema(
+            node_id="BriaVideoGreenScreen",
+            display_name="Bria Video Green Screen",
+            category="partner/video/Bria",
+            description="Replace a video's background with a solid chroma-key screen using Bria.",
+            inputs=[
+                IO.Video.Input("video"),
+                IO.Combo.Input(
+                    "green_shade",
+                    options=["broadcast_green", "chroma_green", "blue_screen"],
+                    tooltip="Solid chroma-key shade applied behind the foreground: "
+                    "broadcast_green (#00B140), chroma_green (#00FF00), or blue_screen (#0000FF).",
+                ),
+                IO.Int.Input(
+                    "seed",
+                    default=0,
+                    min=0,
+                    max=2147483647,
+                    display_mode=IO.NumberDisplay.number,
+                    control_after_generate=True,
+                    tooltip="Seed controls whether the node should re-run; "
+                    "results are non-deterministic regardless of seed.",
+                ),
+            ],
+            outputs=[IO.Video.Output()],
+            hidden=[
+                IO.Hidden.auth_token_comfy_org,
+                IO.Hidden.api_key_comfy_org,
+                IO.Hidden.unique_id,
+            ],
+            is_api_node=True,
+            price_badge=IO.PriceBadge(
+                expr="""{"type":"usd","usd":0.0042,"format":{"suffix":"/second"}}""",
+            ),
+        )
+
+    @classmethod
+    async def execute(
+        cls,
+        video: Input.Video,
+        green_shade: str,
+        seed: int,
+    ) -> IO.NodeOutput:
+        validate_video_duration(video, max_duration=60.0)
+        response = await sync_op(
+            cls,
+            ApiEndpoint(path="/proxy/bria/v2/video/edit/green_screen", method="POST"),
+            data=BriaVideoGreenScreenRequest(
+                video=await upload_video_to_comfyapi(cls, video),
+                green_shade=green_shade,
+                output_container_and_codec="mp4_h264",
+                seed=seed,
+            ),
+            response_model=BriaStatusResponse,
+        )
+        response = await poll_op(
+            cls,
+            ApiEndpoint(path=f"/proxy/bria/v2/status/{response.request_id}"),
+            status_extractor=lambda r: r.status,
+            response_model=BriaRemoveVideoBackgroundResponse,
+        )
+        return IO.NodeOutput(await download_url_to_video_output(response.result.video_url))
+
+
+class BriaVideoReplaceBackground(IO.ComfyNode):
+
+    @classmethod
+    def define_schema(cls):
+        return IO.Schema(
+            node_id="BriaVideoReplaceBackground",
+            display_name="Bria Video Replace Background",
+            category="partner/video/Bria",
+            description="Replace a video's background with a supplied image or video using Bria. "
+            "The output keeps the foreground's resolution and frame rate; a background with a "
+            "different aspect ratio is stretched to fit, so match it for undistorted results.",
+            inputs=[
+                IO.Video.Input("video", tooltip="Foreground video whose background is replaced."),
+                IO.Image.Input(
+                    "background_image",
+                    optional=True,
+                    tooltip="Background image to composite behind the foreground. "
+                    "Provide either a background image or a background video, not both.",
+                ),
+                IO.Video.Input(
+                    "background_video",
+                    optional=True,
+                    tooltip="Background video to composite behind the foreground. "
+                    "Provide either a background image or a background video, not both.",
+                ),
+                IO.Int.Input(
+                    "seed",
+                    default=0,
+                    min=0,
+                    max=2147483647,
+                    display_mode=IO.NumberDisplay.number,
+                    control_after_generate=True,
+                    tooltip="Seed controls whether the node should re-run; "
+                    "results are non-deterministic regardless of seed.",
+                ),
+            ],
+            outputs=[IO.Video.Output()],
+            hidden=[
+                IO.Hidden.auth_token_comfy_org,
+                IO.Hidden.api_key_comfy_org,
+                IO.Hidden.unique_id,
+            ],
+            is_api_node=True,
+            price_badge=IO.PriceBadge(
+                expr="""{"type":"usd","usd":0.0042,"format":{"suffix":"/second"}}""",
+            ),
+        )
+
+    @classmethod
+    async def execute(
+        cls,
+        video: Input.Video,
+        seed: int,
+        background_image: Input.Image | None = None,
+        background_video: Input.Video | None = None,
+    ) -> IO.NodeOutput:
+        if (background_image is None) == (background_video is None):
+            raise ValueError("Provide either a background image or a background video, not both.")
+        validate_video_duration(video, max_duration=60.0)
+        if background_video is not None:
+            validate_video_duration(background_video, max_duration=60.0)
+            background_url = await upload_video_to_comfyapi(cls, background_video, wait_label="Uploading background")
+        else:
+            # Bria's replace_background 500s on RGBA, so drop the alpha channel before upload.
+            background_url = await upload_image_to_comfyapi(
+                cls, background_image[:, :, :, :3], wait_label="Uploading background"
+            )
+        response = await sync_op(
+            cls,
+            ApiEndpoint(path="/proxy/bria/v2/video/edit/replace_background", method="POST"),
+            data=BriaVideoReplaceBackgroundRequest(
+                video=await upload_video_to_comfyapi(cls, video),
+                background_url=background_url,
+                output_container_and_codec="mp4_h264",
+                seed=seed,
+            ),
+            response_model=BriaStatusResponse,
+        )
+        response = await poll_op(
+            cls,
+            ApiEndpoint(path=f"/proxy/bria/v2/status/{response.request_id}"),
+            status_extractor=lambda r: r.status,
+            response_model=BriaRemoveVideoBackgroundResponse,
+        )
+        return IO.NodeOutput(await download_url_to_video_output(response.result.video_url))
+
+
+def _video_to_images_and_mask(video: Input.Video) -> tuple[Input.Image, Input.Mask]:
+    """Decode a transparent webm (VP9 + alpha) into image frames and an alpha mask.
+
+    VP9 keeps its alpha in a side layer that PyAV's default vp9 decoder drops, so the frames
+    are decoded with libvpx-vp9. Returns RGB images [B,H,W,3] in 0..1 and a mask [B,H,W]
+    following the Load Image convention (1 = transparent) for compositing or Save WEBM.
+    """
+    rgb_frames: list[torch.Tensor] = []
+    alpha_frames: list[torch.Tensor] = []
+    with av.open(video.get_stream_source(), mode="r") as container:
+        stream = container.streams.video[0]
+        decoder = CodecContext.create("libvpx-vp9", "r") if stream.codec_context.name == "vp9" else None
+        for packet in container.demux(stream):
+            for frame in (decoder.decode(packet) if decoder is not None else packet.decode()):
+                rgba = torch.from_numpy(frame.to_ndarray(format="rgba")).float() / 255.0
+                rgb_frames.append(rgba[..., :3])
+                alpha_frames.append(rgba[..., 3])
+    images = torch.stack(rgb_frames) if rgb_frames else torch.zeros(0, 0, 0, 3)
+    mask = (1.0 - torch.stack(alpha_frames)) if alpha_frames else torch.zeros((images.shape[0], 64, 64))
+    return images, mask
+
+
+class BriaTransparentVideoBackground(IO.ComfyNode):
+
+    @classmethod
+    def define_schema(cls):
+        return IO.Schema(
+            node_id="BriaTransparentVideoBackground",
+            display_name="Bria Remove Video Background (Transparent)",
+            category="partner/video/Bria",
+            description="Remove the background from a video using Bria and return the cut-out frames "
+            "plus an alpha mask. Connect both to a compositing node, or feed them to Save WEBM to "
+            "write a transparent video.",
+            inputs=[
+                IO.Video.Input("video"),
+                IO.Int.Input(
+                    "seed",
+                    default=0,
+                    min=0,
+                    max=2147483647,
+                    display_mode=IO.NumberDisplay.number,
+                    control_after_generate=True,
+                    tooltip="Seed controls whether the node should re-run; "
+                    "results are non-deterministic regardless of seed.",
+                ),
+            ],
+            outputs=[
+                IO.Image.Output(display_name="images"),
+                IO.Mask.Output(display_name="mask"),
+            ],
+            hidden=[
+                IO.Hidden.auth_token_comfy_org,
+                IO.Hidden.api_key_comfy_org,
+                IO.Hidden.unique_id,
+            ],
+            is_api_node=True,
+            price_badge=IO.PriceBadge(
+                expr="""{"type":"usd","usd":0.0042,"format":{"suffix":"/second"}}""",
+            ),
+        )
+
+    @classmethod
+    async def execute(
+        cls,
+        video: Input.Video,
+        seed: int,
+    ) -> IO.NodeOutput:
+        validate_video_duration(video, max_duration=60.0)
+        response = await sync_op(
+            cls,
+            ApiEndpoint(path="/proxy/bria/v2/video/edit/remove_background", method="POST"),
+            data=BriaRemoveVideoBackgroundRequest(
+                video=await upload_video_to_comfyapi(cls, video),
+                background_color="Transparent",
+                output_container_and_codec="webm_vp9",
+                seed=seed,
+            ),
+            response_model=BriaStatusResponse,
+        )
+        response = await poll_op(
+            cls,
+            ApiEndpoint(path=f"/proxy/bria/v2/status/{response.request_id}"),
+            status_extractor=lambda r: r.status,
+            response_model=BriaRemoveVideoBackgroundResponse,
+        )
+        video_out = await download_url_to_video_output(response.result.video_url)
+        images, mask = _video_to_images_and_mask(video_out)
+        return IO.NodeOutput(images, mask)
+
+
 class BriaExtension(ComfyExtension):
     @override
     async def get_node_list(self) -> list[type[IO.ComfyNode]]:
@@ -323,6 +573,9 @@ class BriaExtension(ComfyExtension):
             BriaImageEditNode,
             BriaRemoveImageBackground,
             BriaRemoveVideoBackground,
+            BriaVideoGreenScreen,
+            BriaVideoReplaceBackground,
+            BriaTransparentVideoBackground,
         ]
 
 

comfy_api_nodes/nodes_bytedance.py

@@ -7,6 +7,7 @@ from io import BytesIO
 import torch
 from typing_extensions import override
 
+from comfy.utils import common_upscale
 from comfy_api.latest import IO, ComfyExtension, Input, Types
 from comfy_api_nodes.apis.bytedance import (
     RECOMMENDED_PRESETS,
@@ -131,6 +132,44 @@ def _prepare_seedance_image(image: Input.Image) -> Input.Image:
     return image
 
 
+# Supported output aspect ratios, used to pre-size FLF frames to matching pixel pair to avoid the 1080p stretch jump.
+SEEDANCE2_RATIO_WH = {
+    "16:9": (16, 9),
+    "4:3": (4, 3),
+    "1:1": (1, 1),
+    "3:4": (3, 4),
+    "9:16": (9, 16),
+    "21:9": (21, 9),
+}
+SEEDANCE2_RES_SHORT_SIDE = {"480p": 480, "720p": 720, "1080p": 1080}
+
+
+def _seedance2_target_dims(resolution: str, ratio: str, image: torch.Tensor) -> tuple[int, int]:
+    """Exact supported output (width, height) for (resolution, ratio).
+
+    The shorter side equals the resolution number (e.g. 1080p 16:9 -> 1920x1080). For ratio
+    "adaptive" (or any unexpected value) the ratio is derived from the image's own aspect, snapped
+    to the nearest supported ratio, so the output keeps the frame's orientation.
+    """
+    short = SEEDANCE2_RES_SHORT_SIDE[resolution]
+    if ratio not in SEEDANCE2_RATIO_WH:
+        aspect = image.shape[-2] / image.shape[-3]  # W / H; tensor is (B, H, W, C)
+        ratio = min(SEEDANCE2_RATIO_WH, key=lambda k: abs(SEEDANCE2_RATIO_WH[k][0] / SEEDANCE2_RATIO_WH[k][1] - aspect))
+    rw, rh = SEEDANCE2_RATIO_WH[ratio]
+    if rw >= rh:  # landscape or square: shorter side is the height
+        out_w, out_h = round(short * rw / rh), short
+    else:  # portrait: shorter side is the width
+        out_w, out_h = short, round(short * rh / rw)
+    return out_w - out_w % 2, out_h - out_h % 2
+
+
+def _resize_to_exact(image: torch.Tensor, width: int, height: int) -> torch.Tensor:
+    """Center-crop to the target aspect and resize to exactly width x height (lanczos)."""
+    samples = image.movedim(-1, 1)  # (B, H, W, C) -> (B, C, H, W)
+    resized = common_upscale(samples, width, height, "lanczos", "center")
+    return resized.movedim(1, -1)
+
+
 async def _resolve_reference_assets(
     cls: type[IO.ComfyNode],
     asset_ids: list[str],
@@ -368,7 +407,7 @@ class ByteDanceImageNode(IO.ComfyNode):
         return IO.Schema(
             node_id="ByteDanceImageNode",
             display_name="ByteDance Image",
-            category="image/partner/ByteDance",
+            category="partner/image/ByteDance",
             description="Generate images using ByteDance models via api based on prompt",
             inputs=[
                 IO.Combo.Input("model", options=["seedream-3-0-t2i-250415"]),
@@ -492,7 +531,7 @@ class ByteDanceSeedreamNode(IO.ComfyNode):
         return IO.Schema(
             node_id="ByteDanceSeedreamNode",
             display_name="ByteDance Seedream 4.5 & 5.0",
-            category="image/partner/ByteDance",
+            category="partner/image/ByteDance",
             description="Unified text-to-image generation and precise single-sentence editing at up to 4K resolution.",
             inputs=[
                 IO.Combo.Input(
@@ -754,7 +793,7 @@ class ByteDanceSeedreamNodeV2(IO.ComfyNode):
         return IO.Schema(
             node_id="ByteDanceSeedreamNodeV2",
             display_name="ByteDance Seedream 4.5 & 5.0",
-            category="image/partner/ByteDance",
+            category="partner/image/ByteDance",
             description="Unified text-to-image generation and precise single-sentence editing at up to 4K resolution.",
             inputs=[
                 IO.String.Input(
@@ -920,7 +959,7 @@ class ByteDanceTextToVideoNode(IO.ComfyNode):
         return IO.Schema(
             node_id="ByteDanceTextToVideoNode",
             display_name="ByteDance Text to Video",
-            category="video/partner/ByteDance",
+            category="partner/video/ByteDance",
             description="Generate video using ByteDance models via api based on prompt",
             inputs=[
                 IO.Combo.Input(
@@ -1048,7 +1087,7 @@ class ByteDanceImageToVideoNode(IO.ComfyNode):
         return IO.Schema(
             node_id="ByteDanceImageToVideoNode",
             display_name="ByteDance Image to Video",
-            category="video/partner/ByteDance",
+            category="partner/video/ByteDance",
             description="Generate video using ByteDance models via api based on image and prompt",
             inputs=[
                 IO.Combo.Input(
@@ -1185,7 +1224,7 @@ class ByteDanceFirstLastFrameNode(IO.ComfyNode):
         return IO.Schema(
             node_id="ByteDanceFirstLastFrameNode",
             display_name="ByteDance First-Last-Frame to Video",
-            category="video/partner/ByteDance",
+            category="partner/video/ByteDance",
             description="Generate video using prompt and first and last frames.",
             inputs=[
                 IO.Combo.Input(
@@ -1333,7 +1372,7 @@ class ByteDanceImageReferenceNode(IO.ComfyNode):
         return IO.Schema(
             node_id="ByteDanceImageReferenceNode",
             display_name="ByteDance Reference Images to Video",
-            category="video/partner/ByteDance",
+            category="partner/video/ByteDance",
             description="Generate video using prompt and reference images.",
             inputs=[
                 IO.Combo.Input(
@@ -1576,7 +1615,7 @@ class ByteDance2TextToVideoNode(IO.ComfyNode):
         return IO.Schema(
             node_id="ByteDance2TextToVideoNode",
             display_name="ByteDance Seedance 2.0 Text to Video",
-            category="video/partner/ByteDance",
+            category="partner/video/ByteDance",
             description="Generate video using Seedance 2.0 models based on a text prompt.",
             inputs=[
                 IO.DynamicCombo.Input(
@@ -1677,7 +1716,7 @@ class ByteDance2FirstLastFrameNode(IO.ComfyNode):
         return IO.Schema(
             node_id="ByteDance2FirstLastFrameNode",
             display_name="ByteDance Seedance 2.0 First-Last-Frame to Video",
-            category="video/partner/ByteDance",
+            category="partner/video/ByteDance",
             description="Generate video using Seedance 2.0 from a first frame image and optional last frame image.",
             inputs=[
                 IO.DynamicCombo.Input(
@@ -1790,10 +1829,28 @@ class ByteDance2FirstLastFrameNode(IO.ComfyNode):
         if last_frame is not None and last_frame_asset_id:
             raise ValueError("Provide only one of last_frame or last_frame_asset_id, not both.")
 
-        if first_frame is not None:
-            first_frame = _prepare_seedance_image(first_frame)
-        if last_frame is not None:
-            last_frame = _prepare_seedance_image(last_frame)
+        request_ratio = model["ratio"]
+        if first_frame_asset_id or last_frame_asset_id:
+            if first_frame is not None:
+                first_frame = _prepare_seedance_image(first_frame)
+            if last_frame is not None:
+                last_frame = _prepare_seedance_image(last_frame)
+        else:
+            # The 1080p FLF stretch fix (pre-size frames to a supported pixel pair + submit ratio="adaptive")
+            # only applies to local image inputs we can resize.
+            request_ratio = "adaptive"
+            target_dims: tuple[int, int] | None = None
+            if first_frame is not None:
+                validate_image_aspect_ratio(first_frame, (2, 5), (5, 2), strict=False)  # 0.4 to 2.5
+                validate_image_dimensions(first_frame, min_width=300, min_height=300)
+                target_dims = _seedance2_target_dims(model["resolution"], model["ratio"], first_frame)
+                first_frame = _resize_to_exact(first_frame, *target_dims)
+            if last_frame is not None:
+                validate_image_aspect_ratio(last_frame, (2, 5), (5, 2), strict=False)  # 0.4 to 2.5
+                validate_image_dimensions(last_frame, min_width=300, min_height=300)
+                if target_dims is None:
+                    target_dims = _seedance2_target_dims(model["resolution"], model["ratio"], last_frame)
+                last_frame = _resize_to_exact(last_frame, *target_dims)
 
         asset_ids_to_resolve = [a for a in (first_frame_asset_id, last_frame_asset_id) if a]
         image_assets: dict[str, str] = {}
@@ -1844,7 +1901,7 @@ class ByteDance2FirstLastFrameNode(IO.ComfyNode):
                 content=content,
                 generate_audio=model["generate_audio"],
                 resolution=model["resolution"],
-                ratio=model["ratio"],
+                ratio=request_ratio,
                 duration=model["duration"],
                 seed=seed,
                 watermark=watermark,
@@ -1944,7 +2001,7 @@ class ByteDance2ReferenceNode(IO.ComfyNode):
         return IO.Schema(
             node_id="ByteDance2ReferenceNode",
             display_name="ByteDance Seedance 2.0 Reference to Video",
-            category="video/partner/ByteDance",
+            category="partner/video/ByteDance",
             description="Generate, edit, or extend video using Seedance 2.0 with reference images, "
             "videos, and audio. Supports multimodal reference, video editing, and video extension.",
             inputs=[
@@ -2241,7 +2298,7 @@ class ByteDanceCreateImageAsset(IO.ComfyNode):
         return IO.Schema(
             node_id="ByteDanceCreateImageAsset",
             display_name="ByteDance Create Image Asset",
-            category="image/partner/ByteDance",
+            category="partner/image/ByteDance",
             description=(
                 "Create a Seedance 2.0 personal image asset. Uploads the input image and "
                 "registers it in the given asset group. If group_id is empty, runs a real-person "
@@ -2308,7 +2365,7 @@ class ByteDanceCreateVideoAsset(IO.ComfyNode):
         return IO.Schema(
             node_id="ByteDanceCreateVideoAsset",
             display_name="ByteDance Create Video Asset",
-            category="video/partner/ByteDance",
+            category="partner/video/ByteDance",
             description=(
                 "Create a Seedance 2.0 personal video asset. Uploads the input video and "
                 "registers it in the given asset group. If group_id is empty, runs a real-person "

comfy_api_nodes/nodes_bytedance_llm.py

@@ -144,7 +144,7 @@ class ByteDanceSeedNode(IO.ComfyNode):
         return IO.Schema(
             node_id="ByteDanceSeedNode",
             display_name="ByteDance Seed",
-            category="text/partner/ByteDance",
+            category="partner/text/ByteDance",
             essentials_category="Text Generation",
             description="Generate text responses with ByteDance's Seed 2.0 models. "
             "Provide a text prompt and optionally one or more images or videos for multimodal context.",

comfy_api_nodes/nodes_elevenlabs.py

@@ -69,7 +69,7 @@ class ElevenLabsSpeechToText(IO.ComfyNode):
         return IO.Schema(
             node_id="ElevenLabsSpeechToText",
             display_name="ElevenLabs Speech to Text",
-            category="audio/partner/ElevenLabs",
+            category="partner/audio/ElevenLabs",
             description="Transcribe audio to text. "
             "Supports automatic language detection, speaker diarization, and audio event tagging.",
             inputs=[
@@ -210,7 +210,7 @@ class ElevenLabsVoiceSelector(IO.ComfyNode):
         return IO.Schema(
             node_id="ElevenLabsVoiceSelector",
             display_name="ElevenLabs Voice Selector",
-            category="audio/partner/ElevenLabs",
+            category="partner/audio/ElevenLabs",
             description="Select a predefined ElevenLabs voice for text-to-speech generation.",
             inputs=[
                 IO.Combo.Input(
@@ -239,7 +239,7 @@ class ElevenLabsTextToSpeech(IO.ComfyNode):
         return IO.Schema(
             node_id="ElevenLabsTextToSpeech",
             display_name="ElevenLabs Text to Speech",
-            category="audio/partner/ElevenLabs",
+            category="partner/audio/ElevenLabs",
             description="Convert text to speech.",
             inputs=[
                 IO.Custom(ELEVENLABS_VOICE).Input(
@@ -414,7 +414,7 @@ class ElevenLabsAudioIsolation(IO.ComfyNode):
         return IO.Schema(
             node_id="ElevenLabsAudioIsolation",
             display_name="ElevenLabs Voice Isolation",
-            category="audio/partner/ElevenLabs",
+            category="partner/audio/ElevenLabs",
             description="Remove background noise from audio, isolating vocals or speech.",
             inputs=[
                 IO.Audio.Input(
@@ -459,7 +459,7 @@ class ElevenLabsTextToSoundEffects(IO.ComfyNode):
         return IO.Schema(
             node_id="ElevenLabsTextToSoundEffects",
             display_name="ElevenLabs Text to Sound Effects",
-            category="audio/partner/ElevenLabs",
+            category="partner/audio/ElevenLabs",
             description="Generate sound effects from text descriptions.",
             inputs=[
                 IO.String.Input(
@@ -555,7 +555,7 @@ class ElevenLabsInstantVoiceClone(IO.ComfyNode):
         return IO.Schema(
             node_id="ElevenLabsInstantVoiceClone",
             display_name="ElevenLabs Instant Voice Clone",
-            category="audio/partner/ElevenLabs",
+            category="partner/audio/ElevenLabs",
             description="Create a cloned voice from audio samples. "
             "Provide 1-8 audio recordings of the voice to clone.",
             inputs=[
@@ -658,7 +658,7 @@ class ElevenLabsSpeechToSpeech(IO.ComfyNode):
         return IO.Schema(
             node_id="ElevenLabsSpeechToSpeech",
             display_name="ElevenLabs Speech to Speech",
-            category="audio/partner/ElevenLabs",
+            category="partner/audio/ElevenLabs",
             description="Transform speech from one voice to another while preserving the original content and emotion.",
             inputs=[
                 IO.Custom(ELEVENLABS_VOICE).Input(
@@ -793,7 +793,7 @@ class ElevenLabsTextToDialogue(IO.ComfyNode):
         return IO.Schema(
             node_id="ElevenLabsTextToDialogue",
             display_name="ElevenLabs Text to Dialogue",
-            category="audio/partner/ElevenLabs",
+            category="partner/audio/ElevenLabs",
             description="Generate multi-speaker dialogue from text. Each dialogue entry has its own text and voice.",
             inputs=[
                 IO.Float.Input(