Enable previews by default and over distributed channels

comfy/cli_args.py

@@ -111,7 +111,7 @@ def create_parser() -> argparse.ArgumentParser:
     parser.add_argument("--disable-ipex-optimize", action="store_true",
                         help="Disables ipex.optimize when loading models with Intel GPUs.")
 
-    parser.add_argument("--preview-method", type=LatentPreviewMethod, default=LatentPreviewMethod.NoPreviews,
+    parser.add_argument("--preview-method", type=LatentPreviewMethod, default=LatentPreviewMethod.Auto,
                         help="Default preview method for sampler nodes.", action=EnumAction)
 
     attn_group = parser.add_mutually_exclusive_group()

comfy/cli_args_types.py

@@ -1,6 +1,6 @@
 # Define a class for your command-line arguments
 import enum
-from typing import Optional, List, Callable
+from typing import Optional, List, Callable, Literal
 import configargparse as argparse
 
 ConfigurationExtender = Callable[[argparse.ArgParser], Optional[argparse.ArgParser]]
@@ -50,7 +50,7 @@ class Configuration(dict):
         fp32_text_enc (bool): Use FP32 precision for the text encoder.
         directml (Optional[int]): Use DirectML. -1 for auto-selection.
         disable_ipex_optimize (bool): Disable IPEX optimization for Intel GPUs.
-        preview_method (LatentPreviewMethod): Method for generating previews. Defaults to "none".
+        preview_method (LatentPreviewMethod): Method for generating previews. Defaults to "auto".
         use_split_cross_attention (bool): Use split cross-attention optimization.
         use_quad_cross_attention (bool): Use sub-quadratic cross-attention optimization.
         use_pytorch_cross_attention (bool): Use PyTorch's cross-attention function.
@@ -116,7 +116,7 @@ class Configuration(dict):
         self.fp32_text_enc: bool = False
         self.directml: Optional[int] = None
         self.disable_ipex_optimize: bool = False
-        self.preview_method: str = "none"
+        self.preview_method: LatentPreviewMethod = LatentPreviewMethod.Auto
         self.use_split_cross_attention: bool = False
         self.use_quad_cross_attention: bool = False
         self.use_pytorch_cross_attention: bool = False

comfy/cmd/latent_preview.py

@@ -1,8 +1,11 @@
+from __future__ import annotations
+
 import torch
 from PIL import Image
 import numpy as np
 from ..cli_args import args
 from ..cli_args_types import LatentPreviewMethod
+from ..model_downloader import get_or_download, KNOWN_APPROX_VAES
 from ..taesd.taesd import TAESD
 from ..cmd import folder_paths
 from .. import utils
@@ -59,7 +62,7 @@ def get_previewer(device, latent_format):
                     if fn.startswith(latent_format.taesd_decoder_name)),
                 ""
             )
-            taesd_decoder_path = folder_paths.get_full_path("vae_approx", taesd_decoder_path)
+            taesd_decoder_path = get_or_download("vae_approx", taesd_decoder_path, KNOWN_APPROX_VAES)
 
         if method == LatentPreviewMethod.Auto:
             method = LatentPreviewMethod.Latent2RGB

comfy/cmd/latent_preview_image_encoding.py

@@ -0,0 +1,29 @@
+from __future__ import annotations
+
+import struct
+from io import BytesIO
+from typing import Literal
+
+import PIL.Image
+from PIL import Image, ImageOps
+
+
+def encode_preview_image(image: PIL.Image.Image, image_type: Literal["JPEG", "PNG"], max_size: int):
+    if max_size is not None:
+        if hasattr(Image, 'Resampling'):
+            resampling = Image.Resampling.BILINEAR
+        else:
+            resampling = Image.Resampling.LANCZOS
+
+        image = ImageOps.contain(image, (max_size, max_size), resampling)
+    type_num = 1
+    if image_type == "JPEG":
+        type_num = 1
+    elif image_type == "PNG":
+        type_num = 2
+    bytesIO = BytesIO()
+    header = struct.pack(">I", type_num)
+    bytesIO.write(header)
+    image.save(bytesIO, format=image_type, quality=95, compress_level=1)
+    preview_bytes = bytesIO.getvalue()
+    return preview_bytes

comfy/cmd/server.py

@@ -11,21 +11,24 @@ import sys
 import traceback
 import uuid
 from asyncio import Future, AbstractEventLoop
+from enum import Enum
 from io import BytesIO
 from typing import List, Optional, Dict
 from urllib.parse import quote, urlencode
 from posixpath import join as urljoin
+
 from can_ada import URL, parse as urlparse
 
 import aiofiles
 import aiohttp
-from PIL import Image, ImageOps
+from PIL import Image
 from PIL.PngImagePlugin import PngInfo
 from aiohttp import web
 from pkg_resources import resource_filename
 from typing_extensions import NamedTuple
 
 import comfy.interruption
+from .latent_preview_image_encoding import encode_preview_image
 from .. import model_management
 from .. import utils
 from ..app.user_manager import UserManager
@@ -713,8 +716,15 @@ class PromptServer(ExecutorToClientProgress):
         else:
             await self.send_json(event, data, sid)
 
-    def encode_bytes(self, event, data):
+    def encode_bytes(self, event: int | Enum | str, data):
-        if not isinstance(event, int):
+        # todo: investigate what is propagating these spurious, string-repr'd previews
+        if event == repr(BinaryEventTypes.UNENCODED_PREVIEW_IMAGE):
+            event = BinaryEventTypes.UNENCODED_PREVIEW_IMAGE.value
+        elif event == repr(BinaryEventTypes.PREVIEW_IMAGE):
+            event = BinaryEventTypes.PREVIEW_IMAGE.value
+        elif isinstance(event, Enum):
+            event: int = event.value
+        elif not isinstance(event, int):
             raise RuntimeError(f"Binary event types must be integers, got {event}")
 
         packed = struct.pack(">I", event)
@@ -726,24 +736,7 @@ class PromptServer(ExecutorToClientProgress):
         image_type = image_data[0]
         image = image_data[1]
         max_size = image_data[2]
-        if max_size is not None:
+        preview_bytes = encode_preview_image(image, image_type, max_size)
-            if hasattr(Image, 'Resampling'):
-                resampling = Image.Resampling.BILINEAR
-            else:
-                resampling = Image.Resampling.LANCZOS
-
-            image = ImageOps.contain(image, (max_size, max_size), resampling)
-        type_num = 1
-        if image_type == "JPEG":
-            type_num = 1
-        elif image_type == "PNG":
-            type_num = 2
-
-        bytesIO = BytesIO()
-        header = struct.pack(">I", type_num)
-        bytesIO.write(header)
-        image.save(bytesIO, format=image_type, quality=95, compress_level=1)
-        preview_bytes = bytesIO.getvalue()
         await self.send_bytes(BinaryEventTypes.PREVIEW_IMAGE, preview_bytes, sid=sid)
 
     async def send_bytes(self, event, data, sid=None):

comfy/component_model/executor_types.py

@@ -1,7 +1,8 @@
 from __future__ import annotations  # for Python 3.7-3.9
 
+import PIL.Image
 from typing_extensions import NotRequired, TypedDict
-from typing import Optional, Literal, Protocol, TypeAlias, Union
+from typing import Optional, Literal, Protocol, TypeAlias, Union, NamedTuple
 
 from .queue_types import BinaryEventTypes
 
@@ -34,11 +35,17 @@ class ProgressMessage(TypedDict):
     sid: NotRequired[str]
 
 
+class UnencodedPreviewImageMessage(NamedTuple):
+    format: Literal["JPEG", "PNG"]
+    pil_image: PIL.Image.Image
+    max_size: int = 512
+
+
 ExecutedMessage: TypeAlias = ExecutingMessage
 
 SendSyncEvent: TypeAlias = Union[Literal["status", "executing", "progress", "executed"], BinaryEventTypes, None]
 
-SendSyncData: TypeAlias = Union[StatusMessage, ExecutingMessage, ProgressMessage, bytes, bytearray, None]
+SendSyncData: TypeAlias = Union[StatusMessage, ExecutingMessage, ProgressMessage, UnencodedPreviewImageMessage, bytes, bytearray, str, None]
 
 
 class ExecutorToClientProgress(Protocol):

comfy/distributed/distributed_progress.py

@@ -1,14 +1,16 @@
 from __future__ import annotations
 
 import asyncio
+import base64
 from asyncio import AbstractEventLoop
+from enum import Enum
 from functools import partial
-
 from typing import Optional, Dict, Any
 
 from aio_pika.patterns import RPC
 
-from ..component_model.executor_types import SendSyncEvent, SendSyncData, ExecutorToClientProgress
+from ..component_model.executor_types import SendSyncEvent, SendSyncData, ExecutorToClientProgress, \
+    UnencodedPreviewImageMessage
 from ..component_model.queue_types import BinaryEventTypes
 from ..utils import hijack_progress
 
@@ -17,6 +19,8 @@ async def _progress(event: SendSyncEvent, data: SendSyncData, user_id: Optional[
                     caller_server: Optional[ExecutorToClientProgress] = None) -> None:
     assert caller_server is not None
     assert user_id is not None
+    if event == BinaryEventTypes.PREVIEW_IMAGE or event == BinaryEventTypes.UNENCODED_PREVIEW_IMAGE or isinstance(data, str):
+        data: bytes = base64.b64decode(data)
     caller_server.send_sync(event, data, sid=user_id)
 
 
@@ -39,11 +43,19 @@ class DistributedExecutorToClientProgress(ExecutorToClientProgress):
 
     async def send(self, event: SendSyncEvent, data: SendSyncData, user_id: Optional[str]) -> None:
         # for now, do not send binary data this way, since it cannot be json serialized / it's impractical
-        if event == BinaryEventTypes.PREVIEW_IMAGE or event == BinaryEventTypes.UNENCODED_PREVIEW_IMAGE:
+        if event == BinaryEventTypes.UNENCODED_PREVIEW_IMAGE:
-            return
+            from ..cmd.latent_preview_image_encoding import encode_preview_image
+
+            # encode preview image
+            event = BinaryEventTypes.PREVIEW_IMAGE.value
+            data: UnencodedPreviewImageMessage
+            format, pil_image, max_size = data
+            data: bytes = encode_preview_image(pil_image, format, max_size)
 
         if isinstance(data, bytes) or isinstance(data, bytearray):
-            return
+            if isinstance(event, Enum):
+                event: int = event.value
+            data: str = base64.b64encode(data).decode()
 
         if user_id is None:
             # todo: user_id should never be none here

comfy/distributed/distributed_prompt_queue.py

@@ -15,13 +15,13 @@ from aio_pika.patterns import JsonRPC
 
 from .distributed_progress import ProgressHandlers
 from .distributed_types import RpcRequest, RpcReply
+from .history import History
 from .server_stub import ServerStub
 from ..auth.permissions import jwt_decode
+from ..cmd.server import PromptServer
 from ..component_model.abstract_prompt_queue import AbstractPromptQueue
 from ..component_model.executor_types import ExecutorToClientProgress, SendSyncEvent, SendSyncData
 from ..component_model.queue_types import Flags, HistoryEntry, QueueTuple, QueueItem, ExecutionStatus, TaskInvocation
-from .history import History
-from ..cmd.server import PromptServer
 
 
 class DistributedPromptQueue(AbstractPromptQueue):

comfy/model_downloader.py

@@ -281,6 +281,11 @@ KNOWN_DIFF_CONTROLNETS = [
     HuggingFile("kohya-ss/ControlNet-diff-modules", "diff_control_sd15_seg_fp16.safetensors"),
 ]
 
+KNOWN_APPROX_VAES = [
+    HuggingFile("madebyollin/taesd", "taesd_decoder.safetensors"),
+    HuggingFile("madebyollin/taesdxl", "taesdxl_decoder.safetensors")
+]
+
 
 def add_known_models(folder_name: str, symbol: List[Union[CivitFile, HuggingFile]], *models: Union[CivitFile, HuggingFile]) -> List[Union[CivitFile, HuggingFile]]:
     if args.disable_known_models: