Merge remote-tracking branch 'origin/master' into pixal3d

README.md

@@ -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?

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()
@@ -144,6 +145,7 @@ 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.")
@@ -249,6 +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
 
@@ -258,6 +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/ldm/ideogram4/model.py

@@ -106,11 +106,11 @@ class Ideogram4EmbedScalar(nn.Module):
         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):
+    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(self.mlp_in.weight.dtype)
+        emb = emb.to(dtype)
         emb = F.silu(self.mlp_in(emb))
         return self.mlp_out(emb)
 
@@ -161,7 +161,7 @@ class Ideogram4Transformer(nn.Module):
         x = x * output_image_mask
         h = self.input_proj(x) * output_image_mask
 
-        t_cond = self.t_embedding(t)
+        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))

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
@@ -17,9 +18,7 @@ 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:

comfy/model_base.py

@@ -1837,7 +1837,24 @@ class WAN21_SCAIL2(WAN21_SCAIL):
 
     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 comfy.context_windows.slice_cond(cond_value, window, x_in, device, temporal_dim=2, temporal_offset=1)
+            # 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 just a single frame 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]
+            if full_ref_mask.shape[2] != video_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 + 1].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):

comfy/model_management.py

@@ -534,8 +534,10 @@ 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):
@@ -641,6 +643,8 @@ def free_pins(size, evict_active=False):
     return freed_total
 
 def ensure_pin_budget(size, evict_active=False):
+    if args.high_ram:
+        return True
     if args.fast_disk:
         shortfall = TOTAL_PINNED_MEMORY + size - MAX_PINNED_MEMORY
     else:
@@ -1494,6 +1498,8 @@ 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():

comfy/ops.py

@@ -180,7 +180,7 @@ def cast_modules_with_vbar(comfy_modules, dtype, device, bias_dtype, non_blockin
             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)
             cast_maybe_lowvram_patch(source, pin, offload_stream, xfer_dest2=dest)
@@ -299,21 +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_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
 
@@ -83,6 +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.
@@ -97,6 +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.
@@ -257,6 +270,7 @@ class VideoFromFile(VideoInput):
 
         image_format = 'gbrpf32le'
         process_image_format = lambda a: a
+        align_graph = None
         audio = None
 
         streams = [video_stream]
@@ -310,7 +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()
@@ -377,25 +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
@@ -451,8 +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(
@@ -461,18 +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
@@ -488,10 +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
@@ -505,9 +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

@@ -1400,7 +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
@@ -1413,6 +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.'''
@@ -1429,6 +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
@@ -1451,6 +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
@@ -1654,6 +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:

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/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_bria.py

@@ -289,7 +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"}}""",
             ),
         )
 
@@ -357,7 +357,7 @@ class BriaVideoGreenScreen(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"}}""",
             ),
         )
 
@@ -433,7 +433,7 @@ class BriaVideoReplaceBackground(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"}}""",
             ),
         )
 
@@ -452,7 +452,10 @@ class BriaVideoReplaceBackground(IO.ComfyNode):
             validate_video_duration(background_video, max_duration=60.0)
             background_url = await upload_video_to_comfyapi(cls, background_video, wait_label="Uploading background")
         else:
-            background_url = await upload_image_to_comfyapi(cls, background_image, wait_label="Uploading background")
+            # 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"),
@@ -530,7 +533,7 @@ class BriaTransparentVideoBackground(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"}}""",
             ),
         )
 
@@ -571,7 +574,7 @@ class BriaExtension(ComfyExtension):
             BriaRemoveImageBackground,
             BriaRemoveVideoBackground,
             BriaVideoGreenScreen,
-            # BriaVideoReplaceBackground,  # server returns Status 500 when we pass background video
+            BriaVideoReplaceBackground,
             BriaTransparentVideoBackground,
         ]
 

comfy_api_nodes/nodes_kling.py

@@ -436,7 +436,7 @@ async def execute_text2video(
             negative_prompt=negative_prompt if negative_prompt else None,
             duration=KlingVideoGenDuration(duration),
             mode=KlingVideoGenMode(model_mode),
-            model_name=KlingVideoGenModelName(model_name),
+            model_name=model_name,
             cfg_scale=cfg_scale,
             aspect_ratio=KlingVideoGenAspectRatio(aspect_ratio),
             camera_control=camera_control,

comfy_api_nodes/nodes_openai.py

@@ -9,6 +9,7 @@ from PIL import Image
 from typing_extensions import override
 
 import folder_paths
+from comfy.utils import common_upscale
 from comfy_api.latest import IO, ComfyExtension, Input
 from comfy_api_nodes.apis.openai import (
     InputFileContent,
@@ -62,7 +63,8 @@ async def validate_and_cast_response(response, timeout: int = None) -> torch.Ten
         timeout: Request timeout in seconds. Defaults to None (no timeout).
 
     Returns:
-        A torch.Tensor representing the image (1, H, W, C).
+        A torch.Tensor of shape (N, H, W, C) with all returned images; images whose
+        dimensions differ from the first image's are resized to match it.
 
     Raises:
         ValueError: If the response is not valid.
@@ -89,6 +91,14 @@ async def validate_and_cast_response(response, timeout: int = None) -> torch.Ten
         arr = np.asarray(pil_img).astype(np.float32) / 255.0
         image_tensors.append(torch.from_numpy(arr))
 
+    # With size="auto" the API can return images whose dimensions differ by a few pixels within a single response
+    # resize them to the first image's dimensions so they can be stacked into one batch.
+    ref_h, ref_w = image_tensors[0].shape[:2]
+    for i, t in enumerate(image_tensors):
+        if t.shape[:2] != (ref_h, ref_w):
+            samples = t.unsqueeze(0).movedim(-1, 1)
+            samples = common_upscale(samples, ref_w, ref_h, "bilinear", "center")
+            image_tensors[i] = samples.movedim(1, -1).squeeze(0)
     return torch.stack(image_tensors, dim=0)
 
 

comfy_api_nodes/nodes_runway.py

@@ -30,13 +30,33 @@ from comfy_api_nodes.apis.runway import (
     Model4,
     ReferenceImage,
     RunwayTextToImageAspectRatioEnum,
+    RunwayAleph2IO,
+    RunwayAleph2KeyframeChain,
+    RunwayAleph2KeyframeItem,
+    RunwayAleph2PromptImageChain,
+    RunwayAleph2PromptImageItem,
+    RunwayAleph2Request,
+    RunwayAleph2Response,
+    RunwayAleph2KeyframeSeconds,
+    RunwayAleph2KeyframeAt,
+    RunwayAleph2PromptImage,
+    RunwayAleph2TimestampPosition,
+    RunwayAleph2RelativePosition,
+    RunwayAleph2ContentModeration,
+    KEYFRAME_MODE_SECONDS,
+    KEYFRAME_MODE_AT,
+    PROMPT_IMAGE_MODE_TIMESTAMP,
+    PROMPT_IMAGE_MODE_POSITION,
 )
 from comfy_api_nodes.util import (
     image_tensor_pair_to_batch,
     validate_string,
     validate_image_dimensions,
     validate_image_aspect_ratio,
+    validate_video_duration,
     upload_images_to_comfyapi,
+    upload_image_to_comfyapi,
+    upload_video_to_comfyapi,
     download_url_to_video_output,
     download_url_to_image_tensor,
     ApiEndpoint,
@@ -45,6 +65,7 @@ from comfy_api_nodes.util import (
 )
 
 PATH_IMAGE_TO_VIDEO = "/proxy/runway/image_to_video"
+PATH_VIDEO_TO_VIDEO = "/proxy/runway/video_to_video"
 PATH_TEXT_TO_IMAGE = "/proxy/runway/text_to_image"
 PATH_GET_TASK_STATUS = "/proxy/runway/tasks"
 
@@ -53,12 +74,6 @@ AVERAGE_DURATION_FLF_SECONDS = 256
 AVERAGE_DURATION_T2I_SECONDS = 41
 
 
-class RunwayApiError(Exception):
-    """Base exception for Runway API errors."""
-
-    pass
-
-
 class RunwayGen4TurboAspectRatio(str, Enum):
     """Aspect ratios supported for Image to Video API when using gen4_turbo model."""
 
@@ -84,14 +99,6 @@ def get_video_url_from_task_status(response: TaskStatusResponse) -> str | None:
     return None
 
 
-def extract_progress_from_task_status(
-    response: TaskStatusResponse,
-) -> float | None:
-    if hasattr(response, "progress") and response.progress is not None:
-        return response.progress * 100
-    return None
-
-
 def get_image_url_from_task_status(response: TaskStatusResponse) -> str | None:
     """Returns the image URL from the task status response if it exists."""
     if hasattr(response, "output") and len(response.output) > 0:
@@ -102,14 +109,13 @@ def get_image_url_from_task_status(response: TaskStatusResponse) -> str | None:
 async def get_response(
     cls: type[IO.ComfyNode], task_id: str, estimated_duration: int | None = None
 ) -> TaskStatusResponse:
-    """Poll the task status until it is finished then get the response."""
     return await poll_op(
         cls,
         ApiEndpoint(path=f"{PATH_GET_TASK_STATUS}/{task_id}"),
         response_model=TaskStatusResponse,
-        status_extractor=lambda r: r.status.value,
+        status_extractor=lambda r: r.status,
         estimated_duration=estimated_duration,
-        progress_extractor=extract_progress_from_task_status,
+        progress_extractor=lambda r: r.progress * 100 if r.progress is not None else None,
     )
 
 
@@ -127,7 +133,7 @@ async def generate_video(
 
     final_response = await get_response(cls, initial_response.id, estimated_duration)
     if not final_response.output:
-        raise RunwayApiError("Runway task succeeded but no video data found in response.")
+        raise ValueError("Runway task succeeded but no video data found in response.")
 
     video_url = get_video_url_from_task_status(final_response)
     return await download_url_to_video_output(video_url)
@@ -410,7 +416,7 @@ class RunwayFirstLastFrameNode(IO.ComfyNode):
             mime_type="image/png",
         )
         if len(download_urls) != 2:
-            raise RunwayApiError("Failed to upload one or more images to comfy api.")
+            raise ValueError("Failed to upload one or more images to comfy api.")
 
         return IO.NodeOutput(
             await generate_video(
@@ -514,11 +520,321 @@ class RunwayTextToImageNode(IO.ComfyNode):
             estimated_duration=AVERAGE_DURATION_T2I_SECONDS,
         )
         if not final_response.output:
-            raise RunwayApiError("Runway task succeeded but no image data found in response.")
+            raise ValueError("Runway task succeeded but no image data found in response.")
 
         return IO.NodeOutput(await download_url_to_image_tensor(get_image_url_from_task_status(final_response)))
 
 
+_TIMING_ABSOLUTE = "Absolute time (seconds)"
+_TIMING_FRACTION = "Fraction of duration (0.0-1.0)"
+
+
+class RunwayAleph2KeyframeNode(IO.ComfyNode):
+
+    @classmethod
+    def define_schema(cls):
+        return IO.Schema(
+            node_id="RunwayAleph2KeyframeNode",
+            display_name="Runway Aleph2 Keyframe",
+            category="partner/video/Runway",
+            description="Anchor a guidance image to a moment of the input (source) video, so Aleph2 "
+            "steers the edit at that point of your footage. Connect this to the 'keyframes' input of "
+            "the Runway Aleph2 Video to Video node; chain several together (up to 5) via the optional "
+            "'keyframes' input below.",
+            inputs=[
+                IO.Image.Input(
+                    "image",
+                    tooltip="The guidance image to apply at the chosen moment of the input video.",
+                ),
+                IO.DynamicCombo.Input(
+                    "timing",
+                    options=[
+                        IO.DynamicCombo.Option(
+                            _TIMING_ABSOLUTE,
+                            [
+                                IO.Float.Input(
+                                    "seconds",
+                                    default=0.0,
+                                    min=0.0,
+                                    max=30.0,
+                                    step=0.1,
+                                    display_mode=IO.NumberDisplay.number,
+                                    tooltip="Time in seconds from start of the input video where this image applies.",
+                                ),
+                            ],
+                        ),
+                        IO.DynamicCombo.Option(
+                            _TIMING_FRACTION,
+                            [
+                                IO.Float.Input(
+                                    "fraction",
+                                    default=0.0,
+                                    min=0.0,
+                                    max=1.0,
+                                    step=0.01,
+                                    display_mode=IO.NumberDisplay.number,
+                                    tooltip="Where in the input video this image applies, "
+                                    "as a fraction of its duration (0.0 = start, 1.0 = end).",
+                                ),
+                            ],
+                        ),
+                    ],
+                    tooltip="How to place this image on the input video's timeline.",
+                ),
+                IO.Custom(RunwayAleph2IO.KEYFRAME).Input(
+                    "keyframes",
+                    optional=True,
+                    tooltip="Optional earlier keyframes to chain with this one.",
+                ),
+            ],
+            outputs=[IO.Custom(RunwayAleph2IO.KEYFRAME).Output(display_name="keyframes")],
+        )
+
+    @classmethod
+    def execute(
+        cls,
+        image: Input.Image,
+        timing: dict,
+        keyframes: RunwayAleph2KeyframeChain | None = None,
+    ) -> IO.NodeOutput:
+        chain = keyframes.clone() if keyframes is not None else RunwayAleph2KeyframeChain()
+        if timing["timing"] == _TIMING_ABSOLUTE:
+            mode, value = KEYFRAME_MODE_SECONDS, float(timing["seconds"])
+        else:
+            mode, value = KEYFRAME_MODE_AT, float(timing["fraction"])
+        chain.add(RunwayAleph2KeyframeItem(image=image, mode=mode, value=value))
+        return IO.NodeOutput(chain)
+
+
+class RunwayAleph2PromptImageNode(IO.ComfyNode):
+
+    @classmethod
+    def define_schema(cls):
+        return IO.Schema(
+            node_id="RunwayAleph2PromptImageNode",
+            display_name="Runway Aleph2 Prompt Image",
+            category="partner/video/Runway",
+            description="Anchor a guidance image to a moment of the output (result) video, to guide what "
+            "the edited video looks like at that point. Connect this to the 'prompt_images' input of the "
+            "Runway Aleph2 Video to Video node; chain several together (up to 5) via the optional "
+            "'prompt_images' input below.",
+            inputs=[
+                IO.Image.Input(
+                    "image",
+                    tooltip="The guidance image to place at the chosen moment of the output video.",
+                ),
+                IO.DynamicCombo.Input(
+                    "position",
+                    options=[
+                        IO.DynamicCombo.Option(
+                            _TIMING_ABSOLUTE,
+                            [
+                                IO.Float.Input(
+                                    "seconds",
+                                    default=0.0,
+                                    min=0.0,
+                                    max=30.0,
+                                    step=0.1,
+                                    display_mode=IO.NumberDisplay.number,
+                                    tooltip="Time in seconds from start of the output video where this image applies.",
+                                ),
+                            ],
+                        ),
+                        IO.DynamicCombo.Option(
+                            _TIMING_FRACTION,
+                            [
+                                IO.Float.Input(
+                                    "fraction",
+                                    default=0.0,
+                                    min=0.0,
+                                    max=1.0,
+                                    step=0.01,
+                                    display_mode=IO.NumberDisplay.number,
+                                    tooltip="Where in the output video this image applies, "
+                                    "as a fraction of its duration (0.0 = start, 1.0 = end).",
+                                ),
+                            ],
+                        ),
+                    ],
+                    tooltip="How to place this image on the output video's timeline.",
+                ),
+                IO.Custom(RunwayAleph2IO.PROMPT_IMAGE).Input(
+                    "prompt_images",
+                    optional=True,
+                    tooltip="Optional earlier prompt images to chain with this one.",
+                ),
+            ],
+            outputs=[IO.Custom(RunwayAleph2IO.PROMPT_IMAGE).Output(display_name="prompt_images")],
+        )
+
+    @classmethod
+    def execute(
+        cls,
+        image: Input.Image,
+        position: dict,
+        prompt_images: RunwayAleph2PromptImageChain | None = None,
+    ) -> IO.NodeOutput:
+        chain = prompt_images.clone() if prompt_images is not None else RunwayAleph2PromptImageChain()
+        if position["position"] == _TIMING_ABSOLUTE:
+            mode, value = PROMPT_IMAGE_MODE_TIMESTAMP, float(position["seconds"])
+        else:
+            mode, value = PROMPT_IMAGE_MODE_POSITION, float(position["fraction"])
+        chain.add(RunwayAleph2PromptImageItem(image=image, mode=mode, value=value))
+        return IO.NodeOutput(chain)
+
+
+class RunwayAleph2VideoToVideoNode(IO.ComfyNode):
+
+    @classmethod
+    def define_schema(cls):
+        return IO.Schema(
+            node_id="RunwayAleph2VideoToVideoNode",
+            display_name="Runway Aleph2 Video to Video",
+            category="partner/video/Runway",
+            description="Edit a video with a text prompt using Runway's Aleph2 model. Aleph2 transforms "
+            "your footage (restyle, relight, add or remove elements, change the viewpoint) while keeping "
+            "the original motion and timing; the output resolution matches the input video, which must be "
+            "2-30 seconds at 30 fps or lower. Optionally steer the edit with either keyframes (anchored to "
+            "the input video) or prompt images (anchored to the output video) - use one or the other, not both.",
+            inputs=[
+                IO.String.Input(
+                    "prompt",
+                    multiline=True,
+                    default="",
+                    tooltip="Describes what should appear in the output (1-1000 characters).",
+                ),
+                IO.Video.Input(
+                    "video",
+                    tooltip="Input video to edit. Must be 2-30 seconds at 30 fps or lower.",
+                ),
+                IO.Int.Input(
+                    "seed",
+                    default=0,
+                    min=0,
+                    max=4294967295,
+                    step=1,
+                    control_after_generate=True,
+                    display_mode=IO.NumberDisplay.number,
+                    tooltip="Random seed for generation",
+                ),
+                IO.Combo.Input(
+                    "public_figure_threshold",
+                    options=["auto", "low"],
+                    default="low",
+                    tooltip="Content moderation for recognizable public figures.",
+                ),
+                IO.Custom(RunwayAleph2IO.KEYFRAME).Input(
+                    "keyframes",
+                    optional=True,
+                    tooltip="Guidance images anchored to the input video, from Aleph2 Keyframe nodes (up to 5). "
+                    "Use keyframes or prompt images, not both.",
+                ),
+                IO.Custom(RunwayAleph2IO.PROMPT_IMAGE).Input(
+                    "prompt_images",
+                    optional=True,
+                    tooltip="Guidance images anchored to the output video, from Aleph2 Prompt Image nodes (up to 5). "
+                    "Use keyframes or prompt images, not both.",
+                ),
+            ],
+            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": 0.4004, "format":{"suffix":"/second"}}""",
+            ),
+        )
+
+    @classmethod
+    async def execute(
+        cls,
+        prompt: str,
+        video: Input.Video,
+        seed: int,
+        public_figure_threshold: str = "low",
+        keyframes: RunwayAleph2KeyframeChain | None = None,
+        prompt_images: RunwayAleph2PromptImageChain | None = None,
+    ) -> IO.NodeOutput:
+        validate_string(prompt, min_length=1, max_length=1000)
+        validate_video_duration(
+            video,
+            min_duration=2.0,
+            max_duration=30.0,
+        )
+        try:
+            fps = float(video.get_frame_rate())
+        except Exception:
+            fps = None
+        if fps is not None and fps > 30.0 + 0.01:
+            raise ValueError(f"Input video frame rate ({fps:.2f} fps) exceeds Aleph2's maximum of 30 fps.")
+
+        if (keyframes and keyframes.items) and (prompt_images and prompt_images.items):
+            raise ValueError("Aleph2 accepts either keyframes or prompt images, not both.")
+
+        video_duration: float | None = None
+        try:
+            video_duration = video.get_duration()
+        except Exception:
+            video_duration = None
+
+        def _check_seconds(value: float, label: str) -> None:
+            if video_duration is not None and value > video_duration + 0.0001:
+                raise ValueError(f"{label} {value:.2f}s exceeds the input video duration ({video_duration:.2f}s).")
+
+        video_url = await upload_video_to_comfyapi(cls, video)
+
+        keyframe_models: list[RunwayAleph2KeyframeSeconds | RunwayAleph2KeyframeAt] = []
+        if keyframes is not None:
+            if len(keyframes.items) > 5:
+                raise ValueError("Aleph2 supports at most 5 keyframes.")
+            for item in keyframes.items:
+                image_url = await upload_image_to_comfyapi(cls, item.image, mime_type="image/png")
+                if item.mode == KEYFRAME_MODE_SECONDS:
+                    _check_seconds(item.value, "Keyframe timestamp")
+                    keyframe_models.append(RunwayAleph2KeyframeSeconds(seconds=item.value, uri=image_url))
+                else:
+                    keyframe_models.append(RunwayAleph2KeyframeAt(at=item.value, uri=image_url))
+
+        prompt_image_models: list[RunwayAleph2PromptImage] = []
+        if prompt_images is not None:
+            if len(prompt_images.items) > 5:
+                raise ValueError("Aleph2 supports at most 5 prompt images.")
+            for item in prompt_images.items:
+                image_url = await upload_image_to_comfyapi(cls, item.image, mime_type="image/png")
+                position: RunwayAleph2TimestampPosition | RunwayAleph2RelativePosition
+                if item.mode == PROMPT_IMAGE_MODE_TIMESTAMP:
+                    _check_seconds(item.value, "Prompt image timestamp")
+                    position = RunwayAleph2TimestampPosition(timestampSeconds=item.value)
+                else:
+                    position = RunwayAleph2RelativePosition(positionPercentage=item.value)
+                prompt_image_models.append(RunwayAleph2PromptImage(position=position, uri=image_url))
+
+        initial_response = await sync_op(
+            cls,
+            endpoint=ApiEndpoint(path=PATH_VIDEO_TO_VIDEO, method="POST"),
+            response_model=RunwayAleph2Response,
+            data=RunwayAleph2Request(
+                promptText=prompt,
+                videoUri=video_url,
+                seed=seed,
+                contentModeration=RunwayAleph2ContentModeration(publicFigureThreshold=public_figure_threshold),
+                keyframes=keyframe_models or None,
+                promptImage=prompt_image_models or None,
+            ),
+        )
+
+        final_response = await get_response(cls, initial_response.id)
+        if not final_response.output:
+            raise ValueError("Runway task succeeded but no video data found in response.")
+
+        return IO.NodeOutput(await download_url_to_video_output(get_video_url_from_task_status(final_response)))
+
+
 class RunwayExtension(ComfyExtension):
     @override
     async def get_node_list(self) -> list[type[IO.ComfyNode]]:
@@ -527,6 +843,9 @@ class RunwayExtension(ComfyExtension):
             RunwayImageToVideoNodeGen3a,
             RunwayImageToVideoNodeGen4,
             RunwayTextToImageNode,
+            RunwayAleph2VideoToVideoNode,
+            RunwayAleph2KeyframeNode,
+            RunwayAleph2PromptImageNode,
         ]
 
 

comfy_api_nodes/nodes_sonilo.py

@@ -16,7 +16,7 @@ from comfy_api_nodes.util import (
 )
 from comfy_api_nodes.util._helpers import (
     default_base_url,
-    get_auth_header,
+    get_comfy_api_headers,
     get_node_id,
     is_processing_interrupted,
 )
@@ -100,8 +100,7 @@ class SoniloTextToMusic(IO.ComfyNode):
             node_id="SoniloTextToMusic",
             display_name="Sonilo Text to Music",
             category="partner/audio/Sonilo",
-            description="Generate music from a text prompt using Sonilo's AI model. "
-            "Leave duration at 0 to let the model infer it from the prompt.",
+            description="Generate music from a text prompt using Sonilo's AI model.",
             inputs=[
                 IO.String.Input(
                     "prompt",
@@ -111,11 +110,10 @@ class SoniloTextToMusic(IO.ComfyNode):
                 ),
                 IO.Int.Input(
                     "duration",
-                    default=0,
-                    min=0,
+                    default=30,
+                    min=1,
                     max=360,
-                    tooltip="Target duration in seconds. Set to 0 to let the model "
-                    "infer the duration from the prompt. Maximum: 6 minutes.",
+                    tooltip="Target duration in seconds. Maximum: 6 minutes.",
                 ),
                 IO.Int.Input(
                     "seed",
@@ -136,13 +134,7 @@ class SoniloTextToMusic(IO.ComfyNode):
             is_api_node=True,
             price_badge=IO.PriceBadge(
                 depends_on=IO.PriceBadgeDepends(widgets=["duration"]),
-                expr="""
-                (
-                  widgets.duration > 0
-                    ? {"type":"usd","usd": 0.005 * widgets.duration}
-                    : {"type":"usd","usd": 0.005, "format":{"suffix":"/second"}}
-                )
-                """,
+                expr='{"type":"usd","usd": 0.0025 * widgets.duration}',
             ),
         )
 
@@ -150,14 +142,13 @@ class SoniloTextToMusic(IO.ComfyNode):
     async def execute(
         cls,
         prompt: str,
-        duration: int = 0,
+        duration: int = 1,
         seed: int = 0,
     ) -> IO.NodeOutput:
-        validate_string(prompt, strip_whitespace=True, min_length=1)
+        validate_string(prompt, strip_whitespace=True, min_length=1, max_length=1000)
         form = aiohttp.FormData()
         form.add_field("prompt", prompt)
-        if duration > 0:
-            form.add_field("duration", str(duration))
+        form.add_field("duration", str(duration))
         audio_bytes = await _stream_sonilo_music(
             cls,
             ApiEndpoint(path="/proxy/sonilo/t2m/generate", method="POST"),
@@ -174,8 +165,7 @@ async def _stream_sonilo_music(
     """POST ``form`` to Sonilo, read the NDJSON stream, and return the first stream's audio bytes."""
     url = urljoin(default_base_url().rstrip("/") + "/", endpoint.path.lstrip("/"))
 
-    headers: dict[str, str] = {}
-    headers.update(get_auth_header(cls))
+    headers = get_comfy_api_headers(cls)
     headers.update(endpoint.headers)
 
     node_id = get_node_id(cls)

comfy_api_nodes/nodes_tripo.py

@@ -1,6 +1,6 @@
 from typing_extensions import override
 
-from comfy_api.latest import IO, ComfyExtension, Input
+from comfy_api.latest import IO, ComfyExtension, Input, Types
 from comfy_api_nodes.apis.tripo import (
     TripoAnimateRetargetRequest,
     TripoAnimateRigRequest,
@@ -8,6 +8,7 @@ from comfy_api_nodes.apis.tripo import (
     TripoFileEmptyReference,
     TripoFileReference,
     TripoImageToModelRequest,
+    TripoImportModelRequest,
     TripoModelVersion,
     TripoMultiviewToModelRequest,
     TripoOrientation,
@@ -21,6 +22,7 @@ from comfy_api_nodes.apis.tripo import (
     TripoTaskType,
     TripoTextToModelRequest,
     TripoTextureModelRequest,
+    TripoTexturePrompt,
     TripoUrlReference,
 )
 from comfy_api_nodes.util import (
@@ -28,6 +30,7 @@ from comfy_api_nodes.util import (
     download_url_to_file_3d,
     poll_op,
     sync_op,
+    upload_3d_model_to_comfyapi,
     upload_images_to_comfyapi,
 )
 
@@ -538,6 +541,14 @@ class TripoTextureNode(IO.ComfyNode):
                     optional=True,
                     advanced=True,
                 ),
+                IO.String.Input(
+                    "texture_prompt",
+                    default="",
+                    multiline=True,
+                    optional=True,
+                    tooltip="Optional text guidance for texturing. Required in practice for imported "
+                    "models (Tripo: Import Model), which carry no source image to infer colors from.",
+                ),
             ],
             outputs=[
                 IO.String.Output(display_name="model_file"),  # for backward compatibility only
@@ -571,6 +582,7 @@ class TripoTextureNode(IO.ComfyNode):
         texture_seed: int | None = None,
         texture_quality: str | None = None,
         texture_alignment: str | None = None,
+        texture_prompt: str = "",
     ) -> IO.NodeOutput:
         response = await sync_op(
             cls,
@@ -583,6 +595,7 @@ class TripoTextureNode(IO.ComfyNode):
                 texture_seed=texture_seed,
                 texture_quality=texture_quality,
                 texture_alignment=texture_alignment,
+                texture_prompt=TripoTexturePrompt(text=texture_prompt.strip()) if texture_prompt.strip() else None,
             ),
         )
         return await poll_until_finished(cls, response, average_duration=80)
@@ -915,6 +928,90 @@ class TripoConversionNode(IO.ComfyNode):
         return await poll_until_finished(cls, response, average_duration=30)
 
 
+class TripoImportModelNode(IO.ComfyNode):
+    """Imports an external 3D model into Tripo, producing a MODEL_TASK_ID for post-processing nodes."""
+
+    SUPPORTED_FORMATS = ("glb", "fbx", "obj", "stl")
+
+    @classmethod
+    def define_schema(cls):
+        return IO.Schema(
+            node_id="TripoImportModelNode",
+            display_name="Tripo: Import Model",
+            category="partner/3d/Tripo",
+            description="Import an external 3D model (e.g. from Rodin, Hunyuan3D or a local file) into Tripo "
+            "to use it with Tripo's post-processing nodes: Texture, Rig, Convert. "
+            "GLB is recommended: textures survive import only when embedded in the file. "
+            "Note that texturing an imported model requires a texture prompt.",
+            inputs=[
+                IO.MultiType.Input(
+                    "model_3d",
+                    types=[IO.File3DGLB, IO.File3DFBX, IO.File3DOBJ, IO.File3DSTL, IO.File3DAny],
+                    tooltip="3D model to import (GLB / FBX / OBJ / STL, up to 150 MB). "
+                    "OBJ and STL files carry no embedded textures.",
+                ),
+            ],
+            outputs=[
+                IO.Custom("MODEL_TASK_ID").Output(display_name="model task_id"),
+            ],
+            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":"text","text":"Free"}""",
+            ),
+        )
+
+    @classmethod
+    async def execute(cls, model_3d: Types.File3D) -> IO.NodeOutput:
+        file_format = (model_3d.format or "").lstrip(".").lower()
+        if file_format == "gltf":
+            raise ValueError(
+                "GLTF (.gltf) references external files and cannot be imported. Export a single-file GLB instead."
+            )
+        if file_format not in cls.SUPPORTED_FORMATS:
+            raise ValueError(
+                f"Unsupported 3D format '{file_format or 'unknown'}'. "
+                f"Tripo import supports: {', '.join(f.upper() for f in cls.SUPPORTED_FORMATS)}."
+            )
+        size = len(model_3d.get_bytes())
+        if size > 150 * 1024 * 1024:
+            raise ValueError(f"Model file is {size / (1024 * 1024):.1f} MB; Tripo import allows up to 150 MB.")
+
+        url = await upload_3d_model_to_comfyapi(cls, model_3d, file_format)
+        response = await sync_op(
+            cls,
+            endpoint=ApiEndpoint(path="/proxy/tripo/v2/openapi/import", method="POST"),
+            response_model=TripoTaskResponse,
+            data=TripoImportModelRequest(url=url, format=file_format),
+        )
+        if response.code != 0:
+            raise RuntimeError(f"Failed to import model: {response.error}")
+
+        task_id = response.data.task_id
+        response_poll = await poll_op(
+            cls,
+            poll_endpoint=ApiEndpoint(path=f"/proxy/tripo/v2/openapi/task/{task_id}"),
+            response_model=TripoTaskResponse,
+            failed_statuses=[
+                TripoTaskStatus.FAILED,
+                TripoTaskStatus.CANCELLED,
+                TripoTaskStatus.UNKNOWN,
+                TripoTaskStatus.BANNED,
+                TripoTaskStatus.EXPIRED,
+            ],
+            status_extractor=lambda x: x.data.status,
+            progress_extractor=lambda x: x.data.progress,
+            estimated_duration=10,
+        )
+        if response_poll.data.status != TripoTaskStatus.SUCCESS:
+            raise RuntimeError(f"Failed to import model: {response_poll}")
+        return IO.NodeOutput(task_id)
+
+
 def _p1_price_expr(*, geometry_credits: int, textured_credits: int, detailed_credits: int) -> str:
     return (
         "("
@@ -1292,6 +1389,7 @@ class TripoExtension(ComfyExtension):
             TripoP1TextToModelNode,
             TripoP1ImageToModelNode,
             TripoP1MultiviewToModelNode,
+            TripoImportModelNode,
             TripoTextureNode,
             TripoRefineNode,
             TripoRigNode,

comfy_api_nodes/util/_helpers.py

@@ -9,6 +9,7 @@ from io import BytesIO
 from yarl import URL
 
 from comfy.cli_args import args
+from comfy.deploy_environment import get_deploy_environment
 from comfy.model_management import processing_interrupted
 from comfy_api.latest import IO
 
@@ -35,6 +36,30 @@ def get_auth_header(node_cls: type[IO.ComfyNode]) -> dict[str, str]:
     return {}
 
 
+def get_usage_source(node_cls: type[IO.ComfyNode]) -> str:
+    """Source of the prompt that triggered this API node.
+
+    Defaults to "comfyui-api" when the submitting client didn't identify itself,
+    i.e. a direct API call to this server.
+    """
+    return node_cls.hidden.comfy_usage_source or "comfyui-api"
+
+
+def get_comfy_api_headers(node_cls: type[IO.ComfyNode]) -> dict[str, str]:
+    """Common headers (auth, deploy environment, usage source) for Comfy API requests.
+
+    Centralizes the shared header set so every Comfy API request sends a consistent
+    set and new shared headers only need to be added in one place. Intended for
+    relative/cloud URLs resolved against ``default_base_url()``; because the result
+    includes auth, callers must not attach it to arbitrary absolute/presigned URLs.
+    """
+    return {
+        **get_auth_header(node_cls),
+        "Comfy-Env": get_deploy_environment(),
+        "Comfy-Usage-Source": get_usage_source(node_cls),
+    }
+
+
 def default_base_url() -> str:
     return getattr(args, "comfy_api_base", "https://api.comfy.org")
 

comfy_api_nodes/util/client.py

@@ -19,12 +19,10 @@ from comfy import utils
 from comfy_api.latest import IO
 from server import PromptServer
 
-from comfy.deploy_environment import get_deploy_environment
-
 from . import request_logger
 from ._helpers import (
     default_base_url,
-    get_auth_header,
+    get_comfy_api_headers,
     get_node_id,
     is_processing_interrupted,
     sleep_with_interrupt,
@@ -645,8 +643,7 @@ async def _request_base(cfg: _RequestConfig, expect_binary: bool):
 
         payload_headers = {"Accept": "*/*"} if expect_binary else {"Accept": "application/json"}
         if not parsed_url.scheme and not parsed_url.netloc:  # is URL relative?
-            payload_headers.update(get_auth_header(cfg.node_cls))
-            payload_headers["Comfy-Env"] = get_deploy_environment()
+            payload_headers.update(get_comfy_api_headers(cfg.node_cls))
         if cfg.endpoint.headers:
             payload_headers.update(cfg.endpoint.headers)
 

comfy_api_nodes/util/download_helpers.py

@@ -17,7 +17,7 @@ from folder_paths import get_output_directory
 from . import request_logger
 from ._helpers import (
     default_base_url,
-    get_auth_header,
+    get_comfy_api_headers,
     is_processing_interrupted,
     sleep_with_interrupt,
     to_aiohttp_url,
@@ -64,7 +64,7 @@ async def download_url_to_bytesio(
         if cls is None:
             raise ValueError("For relative 'cloud' paths, the `cls` parameter is required.")
         url = urljoin(default_base_url().rstrip("/") + "/", url.lstrip("/"))
-        headers = get_auth_header(cls)
+        headers = get_comfy_api_headers(cls)
 
     while True:
         attempt += 1

comfy_execution/asset_enrichment.py

@@ -0,0 +1,66 @@
+"""Enrich executed-node output entries with asset id."""
+import logging
+import os
+
+
+def enrich_output_with_assets(output_ui: dict) -> dict:
+    """Register file-type output entries as assets and inject their ``id``.
+
+    Runs at output-processing time, once per produced output, when
+    --enable-assets is set. Returns a new dict; entries without a resolvable
+    on-disk file path are left unchanged. Errors are caught per-entry so a
+    failure never blocks execution or the other entries.
+    """
+    from comfy.cli_args import args
+    if not args.enable_assets:
+        return output_ui
+
+    import folder_paths
+    from app.assets.services.ingest import register_file_in_place, DependencyMissingError
+
+    enriched = {}
+    for key, entries in output_ui.items():
+        if not isinstance(entries, list):
+            enriched[key] = entries
+            continue
+        new_entries = []
+        for entry in entries:
+            if not isinstance(entry, dict) or "filename" not in entry or "type" not in entry:
+                new_entries.append(entry)
+                continue
+            try:
+                base = folder_paths.get_directory_by_type(entry["type"])
+                if base is None:
+                    new_entries.append(entry)
+                    continue
+                base_abs = os.path.abspath(base)
+                abs_path = os.path.abspath(os.path.join(base_abs, entry.get("subfolder") or "", entry["filename"]))
+                try:
+                    if os.path.commonpath([base_abs, abs_path]) != base_abs:
+                        raise ValueError("escapes base")
+                except ValueError:
+                    logging.warning("Asset enrichment skipped (path escapes base): %s", entry.get("filename"))
+                    new_entries.append(entry)
+                    continue
+                if not os.path.isfile(abs_path):
+                    new_entries.append(entry)
+                    continue
+
+                # Register unconditionally: the file was just produced, and
+                # register_file_in_place re-hashes so an overwritten path can
+                # never carry a stale id.
+                result = register_file_in_place(
+                    abs_path=abs_path,
+                    name=entry["filename"],
+                    tags=[entry["type"]],
+                )
+
+                entry = dict(entry)
+                entry["id"] = result.ref.id
+            except DependencyMissingError:
+                logging.warning("Asset enrichment skipped (blake3 not available): %s", entry.get("filename"))
+            except Exception:
+                logging.warning("Failed to enrich output entry with asset id: %s", entry.get("filename"), exc_info=True)
+            new_entries.append(entry)
+        enriched[key] = new_entries
+    return enriched

comfy_execution/jobs.py

@@ -3,6 +3,7 @@ Job utilities for the /api/jobs endpoint.
 Provides normalization and helper functions for job status tracking.
 """
 
+import uuid
 from typing import Optional
 
 from comfy_api.internal import prune_dict
@@ -19,6 +20,25 @@ class JobStatus:
     ALL = [PENDING, IN_PROGRESS, COMPLETED, FAILED, CANCELLED]
 
 
+def validate_job_id(value) -> str:
+    """Validate a client-supplied job (prompt) id.
+
+    Job ids must be UUIDs in the canonical lowercase hyphenated form. The id
+    is stored and compared verbatim everywhere downstream — history keys,
+    websocket events, and /interrupt matching — so accepting another spelling
+    would silently rewrite the client's id and then miss every exact-match
+    lookup. Rejecting loudly beats that.
+
+    Returns the id unchanged. Raises ValueError when the value is not a
+    string in canonical UUID form.
+    """
+    if not isinstance(value, str):
+        raise ValueError(f"job id must be a string, got {type(value).__name__}")
+    if str(uuid.UUID(value)) != value:
+        raise ValueError("job id must be a UUID in canonical lowercase hyphenated form")
+    return value
+
+
 # Media types that can be previewed in the frontend
 PREVIEWABLE_MEDIA_TYPES = frozenset({'images', 'video', 'audio', '3d', 'text'})
 

comfy_extras/nodes_flux.py

@@ -245,6 +245,11 @@ class KV_Attn_Input:
         cache_key = "{}_{}".format(extra_options["block_type"], extra_options["block_index"])
         if cache_key in self.cache:
             kk, vv = self.cache[cache_key]
+
+            # Fix batch size changing.
+            kk = comfy.utils.repeat_to_batch_size(kk, k.shape[0])
+            vv = comfy.utils.repeat_to_batch_size(vv, v.shape[0])
+
             self.set_cache = False
             return {"q": q, "k": torch.cat((k, kk), dim=2), "v": torch.cat((v, vv), dim=2)}
 

comfy_extras/nodes_rtdetr.py

@@ -14,7 +14,7 @@ class RTDETR_detect(io.ComfyNode):
     def define_schema(cls):
         return io.Schema(
             node_id="RTDETR_detect",
-            display_name="RT-DETR Detect",
+            display_name="Run Real-Time Detection (RT-DETR)",
             category="image/detection",
             search_aliases=["bbox", "bounding box", "object detection", "coco"],
             inputs=[

comfy_extras/nodes_sam3.py

@@ -264,7 +264,7 @@ class SAM3_VideoTrack(io.ComfyNode):
     def define_schema(cls):
         return io.Schema(
             node_id="SAM3_VideoTrack",
-            display_name="SAM3 Video Track",
+            display_name="Run SAM3 Video Track",
             category="image/detection",
             search_aliases=["sam3", "video", "track", "propagate"],
             inputs=[

comfy_extras/nodes_scail.py

@@ -267,7 +267,8 @@ class SCAIL2ColoredMask(io.ComfyNode):
                 io.Combo.Input("sort_by", options=["none", "left_to_right", "area"], default="left_to_right",
                                tooltip="Order in which palette colors are assigned to the tracked objects (applied to both reference and pose video so each identity keeps the same color). left_to_right = leftmost object (by first-frame centroid) gets the first color; area = biggest object (by first-frame mask area) gets the first color; none = keep SAM3's order."),
                 io.Boolean.Input("replacement_mode", default=False,
-                                 tooltip="False = mask_video has black bg (Animation Mode). True = white bg (Replacement Mode). Set the matching replacement_mode on WanSCAILToVideo. reference_image_mask is always black-bg regardless."),
+                                 tooltip="False = Animation Mode (pose_video_mask has black background, reference_image_mask has white background). "
+                                         "True = Replacement Mode (pose_video_mask has white background, reference_image_mask has black background)."),
             ],
             outputs=[
                 io.Image.Output("pose_video_mask"),
@@ -296,14 +297,17 @@ class SCAIL2ColoredMask(io.ComfyNode):
             return td
 
         drv = _prep(driving_track_data)
+        # Animation: driving=black, ref=white. Replacement: driving=white, ref=black.
         mask_video = _render_colored_masks(drv, "white" if replacement_mode else "black")
+        ref_bg = "black" if replacement_mode else "white"
 
         if ref_track_data is not None:
             ref = _prep(ref_track_data)
-            reference_image_mask = _render_colored_masks(ref, "black")
+            reference_image_mask = _render_colored_masks(ref, ref_bg)
         else:
             H, W = drv["orig_size"]
-            reference_image_mask = torch.zeros(1, H, W, 3, device=comfy.model_management.intermediate_device(), dtype=comfy.model_management.intermediate_dtype())
+            fill_value = 1.0 if ref_bg == "white" else 0.0
+            reference_image_mask = torch.full((1, H, W, 3), fill_value, device=comfy.model_management.intermediate_device(), dtype=comfy.model_management.intermediate_dtype())
 
         return io.NodeOutput(mask_video, reference_image_mask)
 

comfy_extras/nodes_sdpose.py

@@ -96,8 +96,12 @@ class KeypointDraw:
         # Body connections - matching DWPose limbSeq (1-indexed, converted to 0-indexed)
         self.body_limbSeq = [
             [2, 3], [2, 6], [3, 4], [4, 5], [6, 7], [7, 8], [2, 9], [9, 10],
-            [10, 11], [2, 12], [12, 13], [13, 14], [2, 1], [1, 15], [15, 17],
-            [1, 16], [16, 18]
+            [10, 11], [2, 12], [12, 13], [13, 14]
+        ]
+
+        # Head connections (1-indexed, converted to 0-indexed)
+        self.head_edges = [
+            [2, 1], [1, 15], [15, 17], [1, 16], [16, 18]
         ]
 
         # Colors matching DWPose
@@ -215,7 +219,7 @@ class KeypointDraw:
         return unique_pts if len(unique_pts) > 1 else [[center[0], center[1]], [center[0], center[1]]]
 
     def draw_wholebody_keypoints(self, canvas, keypoints, scores=None, threshold=0.3,
-                                 draw_body=True, draw_feet=True, draw_face=True, draw_hands=True, stick_width=4, face_point_size=3):
+                                 draw_body=True, draw_head=True, draw_feet=True, draw_face=True, draw_hands=True, stick_width=4, face_point_size=3):
         """
         Draw wholebody keypoints (134 keypoints after processing) in DWPose style.
 
@@ -237,9 +241,17 @@ class KeypointDraw:
         """
         H, W, C = canvas.shape
 
-        # Draw body limbs
-        if draw_body and len(keypoints) >= 18:
-            for i, limb in enumerate(self.body_limbSeq):
+        # Draw body limbs & head connections
+        if (draw_body or draw_head) and len(keypoints) >= 18:
+            colorIndexOffset = 0
+            edges = []
+            if draw_body:
+                edges += self.body_limbSeq
+            else:
+                colorIndexOffset += len(self.body_limbSeq)
+            if draw_head:
+                edges += self.head_edges
+            for i, limb in enumerate(edges):
                 # Convert from 1-indexed to 0-indexed
                 idx1, idx2 = limb[0] - 1, limb[1] - 1
 
@@ -262,11 +274,17 @@ class KeypointDraw:
 
                 polygon = self.draw.ellipse2Poly((int(mY), int(mX)), (int(length / 2), stick_width), int(angle), 0, 360, 1)
 
-                self.draw.fillConvexPoly(canvas, polygon, self.colors[i % len(self.colors)])
+                self.draw.fillConvexPoly(canvas, polygon, self.colors[(i + colorIndexOffset) % len(self.colors)])
 
-        # Draw body keypoints
-        if draw_body and len(keypoints) >= 18:
+        # Draw body & head keypoints
+        if (draw_body or draw_head) and len(keypoints) >= 18:
+            head_keypoints = {0, 14, 15, 16, 17} # nose, eyes, ears
+            neck_point = 1
             for i in range(18):
+                if not draw_head and i in head_keypoints:
+                    continue
+                if not draw_body and i not in head_keypoints and i != neck_point:
+                    continue
                 if scores is not None and scores[i] < threshold:
                     continue
                 x, y = int(keypoints[i][0]), int(keypoints[i][1])
@@ -365,6 +383,7 @@ class SDPoseDrawKeypoints(io.ComfyNode):
                 io.Int.Input("stick_width", default=4, min=1, max=10, step=1),
                 io.Int.Input("face_point_size", default=3, min=1, max=10, step=1),
                 io.Float.Input("score_threshold", default=0.3, min=0.0, max=1.0, step=0.01),
+                io.Boolean.Input("draw_head", default=True),
             ],
             outputs=[
                 io.Image.Output(),
@@ -372,7 +391,7 @@ class SDPoseDrawKeypoints(io.ComfyNode):
         )
 
     @classmethod
-    def execute(cls, keypoints, draw_body, draw_hands, draw_face, draw_feet, stick_width, face_point_size, score_threshold) -> io.NodeOutput:
+    def execute(cls, keypoints, draw_body, draw_hands, draw_face, draw_feet, stick_width, face_point_size, score_threshold, draw_head) -> io.NodeOutput:
         if not keypoints:
             return io.NodeOutput(torch.zeros((1, 64, 64, 3), dtype=torch.float32))
         height = keypoints[0]["canvas_height"]
@@ -405,7 +424,7 @@ class SDPoseDrawKeypoints(io.ComfyNode):
                 canvas = drawer.draw_wholebody_keypoints(
                     canvas, kp, sc,
                     threshold=score_threshold,
-                    draw_body=draw_body, draw_feet=draw_feet,
+                    draw_body=draw_body, draw_head=draw_head, draw_feet=draw_feet,
                     draw_face=draw_face, draw_hands=draw_hands,
                     stick_width=stick_width, face_point_size=face_point_size,
                 )

comfy_extras/nodes_video.py

@@ -134,6 +134,17 @@ class CreateVideo(io.ComfyNode):
                 io.Image.Input("images", tooltip="The images to create a video from."),
                 io.Float.Input("fps", default=30.0, min=1.0, max=120.0, step=1.0),
                 io.Audio.Input("audio", optional=True, tooltip="The audio to add to the video."),
+                io.Int.Input(
+                    "bit_depth",
+                    min=8,
+                    max=10,
+                    default=8,
+                    step=2,
+                    tooltip="Bit depth of the created video. 10-bit keeps smoother gradients with less"
+                    " banding, but some players and downstream nodes may not support it.",
+                    optional=True,
+                    display_mode=io.NumberDisplay.number,
+                ),
             ],
             outputs=[
                 io.Video.Output(),
@@ -141,9 +152,14 @@ class CreateVideo(io.ComfyNode):
         )
 
     @classmethod
-    def execute(cls, images: Input.Image, fps: float, audio: Optional[Input.Audio] = None) -> io.NodeOutput:
+    def execute(
+        cls, images: Input.Image, fps: float, audio: Optional[Input.Audio] = None, bit_depth: int = 8,
+    ) -> io.NodeOutput:
         return io.NodeOutput(
-            InputImpl.VideoFromComponents(Types.VideoComponents(images=images, audio=audio, frame_rate=Fraction(fps)))
+            InputImpl.VideoFromComponents(
+                Types.VideoComponents(images=images, audio=audio, frame_rate=Fraction(fps)),
+                bit_depth=bit_depth,
+            )
         )
 
 class GetVideoComponents(io.ComfyNode):
@@ -154,7 +170,7 @@ class GetVideoComponents(io.ComfyNode):
             search_aliases=["extract frames", "split video", "video to images", "demux"],
             display_name="Get Video Components",
             category="video",
-            description="Extracts all components from a video: frames, audio, and framerate.",
+            description="Extracts all components from a video: frames, audio, framerate, and bit depth.",
             inputs=[
                 io.Video.Input("video", tooltip="The video to extract components from."),
             ],
@@ -162,13 +178,14 @@ class GetVideoComponents(io.ComfyNode):
                 io.Image.Output(display_name="images"),
                 io.Audio.Output(display_name="audio"),
                 io.Float.Output(display_name="fps"),
+                io.Int.Output(display_name="bit_depth"),
             ],
         )
 
     @classmethod
     def execute(cls, video: Input.Video) -> io.NodeOutput:
         components = video.get_components()
-        return io.NodeOutput(components.images, components.audio, float(components.frame_rate))
+        return io.NodeOutput(components.images, components.audio, float(components.frame_rate), video.get_bit_depth())
 
 
 class LoadVideo(io.ComfyNode):

comfyui_version.py

@@ -1,3 +1,3 @@
 # This file is automatically generated by the build process when version is
 # updated in pyproject.toml.
-__version__ = "0.24.0"
+__version__ = "0.25.0"

execution.py

@@ -40,6 +40,7 @@ from comfy_execution.graph_utils import GraphBuilder, is_link
 from comfy_execution.validation import validate_node_input
 from comfy_execution.progress import get_progress_state, reset_progress_state, add_progress_handler, WebUIProgressHandler
 from comfy_execution.utils import CurrentNodeContext
+from comfy_execution.asset_enrichment import enrich_output_with_assets
 from comfy_api.internal import _ComfyNodeInternal, _NodeOutputInternal, first_real_override, is_class, make_locked_method_func
 from comfy_api.latest import io, _io
 from comfy_execution.cache_provider import _has_cache_providers, _get_cache_providers, _logger as _cache_logger
@@ -199,6 +200,8 @@ def get_input_data(inputs, class_def, unique_id, execution_list=None, dynprompt=
                 hidden_inputs_v3[io.Hidden.auth_token_comfy_org] = extra_data.get("auth_token_comfy_org", None)
             if io.Hidden.api_key_comfy_org.name in hidden:
                 hidden_inputs_v3[io.Hidden.api_key_comfy_org] = extra_data.get("api_key_comfy_org", None)
+            if io.Hidden.comfy_usage_source.name in hidden:
+                hidden_inputs_v3[io.Hidden.comfy_usage_source] = extra_data.get("comfy_usage_source", None)
     else:
         if "hidden" in valid_inputs:
             h = valid_inputs["hidden"]
@@ -215,6 +218,8 @@ def get_input_data(inputs, class_def, unique_id, execution_list=None, dynprompt=
                     input_data_all[x] = [extra_data.get("auth_token_comfy_org", None)]
                 if h[x] == "API_KEY_COMFY_ORG":
                     input_data_all[x] = [extra_data.get("api_key_comfy_org", None)]
+                if h[x] == "COMFY_USAGE_SOURCE":
+                    input_data_all[x] = [extra_data.get("comfy_usage_source", None)]
     v3_data["hidden_inputs"] = hidden_inputs_v3
     return input_data_all, missing_keys, v3_data
 
@@ -418,6 +423,7 @@ def _is_intermediate_output(dynprompt, node_id):
     class_def = nodes.NODE_CLASS_MAPPINGS[class_type]
     return getattr(class_def, 'HAS_INTERMEDIATE_OUTPUT', False)
 
+
 def _send_cached_ui(server, node_id, display_node_id, cached, prompt_id, ui_outputs):
     if server.client_id is None:
         return
@@ -552,6 +558,10 @@ async def execute(server, dynprompt, caches, current_item, extra_data, executed,
                 asyncio.create_task(await_completion())
                 return (ExecutionResult.PENDING, None, None)
         if len(output_ui) > 0:
+            # Enrich at output-processing time (not in the send path) so assets
+            # are registered even when no client is connected, and the asset id
+            # flows into ui_outputs and the cache alongside the raw entries.
+            output_ui = enrich_output_with_assets(output_ui)
             ui_outputs[unique_id] = {
                 "meta": {
                     "node_id": unique_id,

main.py

@@ -55,7 +55,11 @@ if __name__ == "__main__" and args.debug_hang:
 import comfy_aimdo.control
 
 if enables_dynamic_vram():
-    comfy_aimdo.control.init()
+    try:
+        comfy_aimdo.control.init(simple_vram_headroom=None if args.reserve_vram is None else int(args.reserve_vram * 1024 ** 3))
+    except TypeError:
+        # comfy-aimdo 0.4.9 protocol.
+        comfy_aimdo.control.init()
 
 if os.name == "nt":
     os.environ['MIMALLOC_PURGE_DELAY'] = '0'
@@ -231,23 +235,30 @@ import comfy.model_patcher
 if args.enable_dynamic_vram or (enables_dynamic_vram() and comfy.model_management.is_nvidia() and not comfy.model_management.is_wsl()):
     if (not args.enable_dynamic_vram) and (comfy.model_management.torch_version_numeric < (2, 8)):
         logging.warning("Unsupported Pytorch detected. DynamicVRAM support requires Pytorch version 2.8 or later. Falling back to legacy ModelPatcher. VRAM estimates may be unreliable especially on Windows")
-    elif comfy_aimdo.control.init_devices(d.index for d in comfy.model_management.get_all_torch_devices()):
-        if args.verbose == 'DEBUG':
-            comfy_aimdo.control.set_log_debug()
-        elif args.verbose == 'CRITICAL':
-            comfy_aimdo.control.set_log_critical()
-        elif args.verbose == 'ERROR':
-            comfy_aimdo.control.set_log_error()
-        elif args.verbose == 'WARNING':
-            comfy_aimdo.control.set_log_warning()
-        else: #INFO
-            comfy_aimdo.control.set_log_info()
-
-        comfy.model_patcher.CoreModelPatcher = comfy.model_patcher.ModelPatcherDynamic
-        comfy.memory_management.aimdo_enabled = True
-        logging.info("DynamicVRAM support detected and enabled")
     else:
-        logging.warning("No working comfy-aimdo install detected. DynamicVRAM support disabled. Falling back to legacy ModelPatcher. VRAM estimates may be unreliable especially on Windows")
+        try:
+            aimdo_initialized = comfy_aimdo.control.init_devices((d.index, int(args.vram_headroom * 1024 ** 3)) for d in comfy.model_management.get_all_torch_devices())
+        except TypeError:
+            # comfy-aimdo 0.4.9 protocol.
+            aimdo_initialized = comfy_aimdo.control.init_devices(d.index for d in comfy.model_management.get_all_torch_devices())
+
+        if aimdo_initialized:
+            if args.verbose == 'DEBUG':
+                comfy_aimdo.control.set_log_debug()
+            elif args.verbose == 'CRITICAL':
+                comfy_aimdo.control.set_log_critical()
+            elif args.verbose == 'ERROR':
+                comfy_aimdo.control.set_log_error()
+            elif args.verbose == 'WARNING':
+                comfy_aimdo.control.set_log_warning()
+            else: #INFO
+                comfy_aimdo.control.set_log_info()
+
+            comfy.model_patcher.CoreModelPatcher = comfy.model_patcher.ModelPatcherDynamic
+            comfy.memory_management.aimdo_enabled = True
+            logging.info("DynamicVRAM support detected and enabled")
+        else:
+            logging.warning("No working comfy-aimdo install detected. DynamicVRAM support disabled. Falling back to legacy ModelPatcher. VRAM estimates may be unreliable especially on Windows")
 
 
 def cuda_malloc_warning():
@@ -490,6 +501,11 @@ def start_comfyui(asyncio_loop=None):
         init_custom_nodes=(not args.disable_all_custom_nodes) or len(args.whitelist_custom_nodes) > 0,
         init_api_nodes=not args.disable_api_nodes
     ))
+
+    # Re-apply Comfy's cuDNN benchmark policy after custom-node imports. Benchmark
+    # mode can request near-card-sized autotune workspaces, and some custom nodes set it at import time.
+    comfy.model_management.set_cudnn_benchmark()
+
     hook_breaker_ac10a0.restore_functions()
 
     cuda_malloc_warning()

manager_requirements.txt

@@ -1 +1 @@
-comfyui_manager==4.2.1
+comfyui_manager==4.2.2

openapi.yaml

@@ -1062,6 +1062,9 @@ components:
                         comfyui_version:
                             description: ComfyUI version
                             type: string
+                        deploy_environment:
+                            description: How this ComfyUI instance is deployed (e.g. cloud, local-git, local-portable, local-desktop)
+                            type: string
                         embedded_python:
                             description: Whether using embedded Python
                             type: boolean
@@ -1792,7 +1795,9 @@ paths:
                         application/json:
                             schema:
                                 $ref: '#/components/schemas/ErrorResponse'
-                    description: Invalid request (no fields provided)
+                    description: |
+                        Invalid request — no fields provided, or `preview_id` is the zero UUID
+                        (`INVALID_PREVIEW_ID`).
                 "401":
                     content:
                         application/json:
@@ -1804,7 +1809,10 @@ paths:
                         application/json:
                             schema:
                                 $ref: '#/components/schemas/ErrorResponse'
-                    description: Asset not found
+                    description: |
+                        Asset not found — returned both when the asset being updated does
+                        not exist and when `preview_id` does not reference an asset
+                        accessible to the caller.
                 "500":
                     content:
                         application/json:
@@ -3042,6 +3050,12 @@ paths:
                             schema:
                                 $ref: '#/components/schemas/PromptErrorResponse'
                     description: Payment required - Insufficient credits
+                "413":
+                    content:
+                        application/json:
+                            schema:
+                                $ref: '#/components/schemas/PromptErrorResponse'
+                    description: Workflow JSON too large
                 "429":
                     content:
                         application/json:

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "ComfyUI"
-version = "0.24.0"
+version = "0.25.0"
 readme = "README.md"
 license = { file = "LICENSE" }
 requires-python = ">=3.10"

requirements.txt

@@ -1,6 +1,6 @@
 comfyui-frontend-package==1.45.15
-comfyui-workflow-templates==0.9.98
-comfyui-embedded-docs==0.5.3
+comfyui-workflow-templates==0.10.0
+comfyui-embedded-docs==0.5.4
 torch
 torchsde
 torchvision
@@ -23,7 +23,7 @@ SQLAlchemy>=2.0.0
 filelock
 av>=16.0.0
 comfy-kitchen==0.2.10
-comfy-aimdo==0.4.9
+comfy-aimdo==0.4.10
 requests
 simpleeval>=1.0.0
 blake3

server.py

@@ -8,7 +8,7 @@ import time
 import nodes
 import folder_paths
 import execution
-from comfy_execution.jobs import JobStatus, get_job, get_all_jobs
+from comfy_execution.jobs import JobStatus, get_job, get_all_jobs, validate_job_id
 import uuid
 import urllib
 import json
@@ -27,6 +27,7 @@ import logging
 
 import mimetypes
 from comfy.cli_args import args
+from comfy.deploy_environment import get_deploy_environment
 import comfy.utils
 import comfy.model_management
 from comfy_api import feature_flags
@@ -690,6 +691,7 @@ class PromptServer():
                     "python_version": sys.version,
                     "pytorch_version": comfy.model_management.torch_version,
                     "embedded_python": os.path.split(os.path.split(sys.executable)[0])[1] == "python_embeded",
+                    "deploy_environment": get_deploy_environment(),
                     "argv": sys.argv
                 },
                 "devices": device_entries
@@ -942,7 +944,21 @@ class PromptServer():
 
             if "prompt" in json_data:
                 prompt = json_data["prompt"]
-                prompt_id = str(json_data.get("prompt_id", uuid.uuid4()))
+                client_prompt_id = json_data.get("prompt_id")
+                if client_prompt_id is None:
+                    # Absent or explicit null: the server mints the id.
+                    prompt_id = str(uuid.uuid4())
+                else:
+                    try:
+                        prompt_id = validate_job_id(client_prompt_id)
+                    except ValueError:
+                        error = {
+                            "type": "invalid_prompt_id",
+                            "message": "prompt_id must be a valid UUID",
+                            "details": "prompt_id must be a UUID string in canonical lowercase hyphenated form; omit it to let the server generate one",
+                            "extra_info": {}
+                        }
+                        return web.json_response({"error": error, "node_errors": {}}, status=400)
 
                 partial_execution_targets = None
                 if "partial_execution_targets" in json_data:
@@ -957,6 +973,11 @@ class PromptServer():
 
                 if "client_id" in json_data:
                     extra_data["client_id"] = json_data["client_id"]
+
+                if "comfy_usage_source" not in extra_data:
+                    usage_source = request.headers.get("Comfy-Usage-Source")
+                    if usage_source:
+                        extra_data["comfy_usage_source"] = usage_source
                 if valid[0]:
                     outputs_to_execute = valid[2]
                     sensitive = {}

tests-unit/assets_test/test_prompt_id_enforcement.py

@@ -0,0 +1,69 @@
+"""POST /prompt enforces canonical-UUID job ids at creation time.
+
+Lives in assets_test because it uses this suite's booted-server fixture. The
+invariant itself is pipeline-wide: a job id is stored and compared verbatim
+downstream — history keys, websocket correlation, and /interrupt matching —
+so a job minted with a non-canonical id would miss every exact-match lookup.
+
+The prompt bodies here are intentionally invalid workflows — prompt_id
+validation happens before workflow validation, so a rejected id returns
+``invalid_prompt_id`` while an accepted id falls through to the ordinary
+workflow-validation error (proving it cleared the id check).
+"""
+import requests
+
+
+def _post_prompt(http: requests.Session, api_base: str, body: dict) -> requests.Response:
+    return http.post(api_base + "/prompt", json=body, timeout=30)
+
+
+def _error_type(r: requests.Response) -> str:
+    return r.json()["error"]["type"]
+
+
+def test_non_uuid_prompt_id_rejected(http: requests.Session, api_base: str):
+    r = _post_prompt(http, api_base, {"prompt": {}, "prompt_id": "not-a-uuid"})
+    assert r.status_code == 400, r.text
+    assert _error_type(r) == "invalid_prompt_id"
+
+
+def test_non_string_prompt_id_rejected(http: requests.Session, api_base: str):
+    # Previously str()-coerced (123 became the job id "123"); must now be a 400,
+    # not a 500 from uuid.UUID choking on a non-string.
+    r = _post_prompt(http, api_base, {"prompt": {}, "prompt_id": 123})
+    assert r.status_code == 400, r.text
+    assert _error_type(r) == "invalid_prompt_id"
+
+
+def test_non_canonical_uuid_rejected(http: requests.Session, api_base: str):
+    # Parseable as a UUID, but not the canonical lowercase form: rejected
+    # loudly rather than silently rewritten (downstream lookups match the
+    # stored id exactly).
+    r = _post_prompt(
+        http,
+        api_base,
+        {"prompt": {}, "prompt_id": "AAAAAAAA-BBBB-4CCC-8DDD-EEEEEEEEEEEE"},
+    )
+    assert r.status_code == 400, r.text
+    assert _error_type(r) == "invalid_prompt_id"
+
+
+def test_canonical_uuid_accepted(http: requests.Session, api_base: str):
+    # The id clears validation; the empty workflow then fails ordinary prompt
+    # validation, proving the request got past the id check.
+    r = _post_prompt(
+        http,
+        api_base,
+        {"prompt": {}, "prompt_id": "aaaaaaaa-bbbb-4ccc-8ddd-eeeeeeeeeeee"},
+    )
+    assert r.status_code == 400, r.text
+    assert _error_type(r) != "invalid_prompt_id"
+
+
+def test_null_prompt_id_not_rejected(http: requests.Session, api_base: str):
+    # Explicit null means "server generates" and must not be rejected as an
+    # invalid id. (The minted id itself is not observable here because the
+    # workflow is invalid; unit tests cover validate_job_id directly.)
+    r = _post_prompt(http, api_base, {"prompt": {}, "prompt_id": None})
+    assert r.status_code == 400, r.text
+    assert _error_type(r) != "invalid_prompt_id"

tests-unit/comfy_api_test/video_bit_depth_test.py

@@ -0,0 +1,93 @@
+import pytest
+import torch
+import av
+import numpy as np
+from fractions import Fraction
+from comfy_api.latest._input_impl.video_types import VideoFromFile, VideoFromComponents
+from comfy_api.latest._util.video_types import VideoComponents
+
+
+@pytest.fixture(scope="module")
+def gradient_components():
+    """Narrow horizontal ramp (0.25..0.30) that needs more than 8 bits to stay smooth"""
+    width, height, frames = 64, 64, 3
+    ramp = torch.linspace(0.25, 0.30, width).view(1, 1, width, 1).expand(frames, height, width, 3)
+    return VideoComponents(images=ramp.contiguous(), frame_rate=Fraction(30))
+
+
+@pytest.fixture(scope="module")
+def src8(gradient_components, tmp_path_factory):
+    """8-bit h264 mp4 (Create Video default)"""
+    path = str(tmp_path_factory.mktemp("video") / "src8.mp4")
+    VideoFromComponents(gradient_components).save_to(path)
+    return path
+
+
+@pytest.fixture(scope="module")
+def src10(gradient_components, tmp_path_factory):
+    """10-bit h264 mp4 (Create Video with bit_depth=10)"""
+    path = str(tmp_path_factory.mktemp("video") / "src10.mp4")
+    VideoFromComponents(gradient_components, bit_depth=10).save_to(path)
+    return path
+
+
+def probe(path):
+    """(codec, pix_fmt, bit_depth) of the first video stream"""
+    with av.open(path) as container:
+        stream = container.streams.video[0]
+        return (stream.codec.name, stream.format.name, max(c.bits for c in stream.format.components))
+
+
+def decoded_levels(path):
+    """Unique tonal levels in the first decoded frame (banding measure)"""
+    with av.open(path) as container:
+        frame = next(container.decode(container.streams.video[0]))
+        return len(np.unique(frame.to_ndarray(format="gbrpf32le")[..., 0]))
+
+
+def video_packet_bytes(path):
+    """Raw video packet payloads; identical to the source's only for a true remux"""
+    with av.open(path) as container:
+        return [bytes(p) for p in container.demux(container.streams.video[0]) if p.size]
+
+
+def test_create_video_bit_depth(src8, src10):
+    """Create Video's bit_depth picks the encoded depth (default 8-bit); 10-bit reduces banding"""
+    assert probe(src8) == ("h264", "yuv420p", 8)
+    assert probe(src10) == ("h264", "yuv420p10le", 10)
+    assert decoded_levels(src10) > 2 * decoded_levels(src8)
+
+
+def test_save_auto_keeps_source_depth(src8, src10, tmp_path):
+    """Save Video (no bit_depth = auto) stream-copies the source, preserving its depth byte-for-byte"""
+    for name, src in [("p8", src8), ("p10", src10)]:
+        path = str(tmp_path / f"{name}.mp4")
+        VideoFromFile(src).save_to(path)
+        assert probe(path) == probe(src)
+        assert video_packet_bytes(path) == video_packet_bytes(src)
+
+
+def test_save_explicit_depth_reencodes(src8, src10, tmp_path):
+    """An explicit bit_depth different from the source forces a re-encode to that depth"""
+    down = str(tmp_path / "down8.mp4")
+    VideoFromFile(src10).save_to(down, bit_depth=8)
+    assert probe(down) == ("h264", "yuv420p", 8)
+
+    up = str(tmp_path / "up10.mp4")
+    VideoFromFile(src8).save_to(up, bit_depth=10)
+    assert probe(up) == ("h264", "yuv420p10le", 10)
+
+
+def test_trim_keeps_source_depth(src10, tmp_path):
+    """Video Slice re-encodes (trim) but preserves the source's 10-bit depth"""
+    path = str(tmp_path / "trim.mp4")
+    VideoFromFile(src10).as_trimmed(start_time=0, duration=1 / 30, strict_duration=False).save_to(path)
+    assert probe(path) == ("h264", "yuv420p10le", 10)
+
+
+def test_get_bit_depth(gradient_components, src8, src10):
+    """get_bit_depth reports a video's depth (backs the Get Video Components output)"""
+    assert VideoFromFile(src8).get_bit_depth() == 8
+    assert VideoFromFile(src10).get_bit_depth() == 10
+    assert VideoFromComponents(gradient_components, bit_depth=10).get_bit_depth() == 10
+    assert VideoFromComponents(gradient_components).get_bit_depth() == 8

tests-unit/execution_test/test_enrich_output.py

@@ -0,0 +1,205 @@
+"""Tests for enrich_output_with_assets in comfy_execution/asset_enrichment.py."""
+import os
+import types
+import unittest
+from unittest.mock import MagicMock, patch
+
+
+def _make_args(enable_assets: bool):
+    a = types.SimpleNamespace()
+    a.enable_assets = enable_assets
+    return a
+
+
+def _make_register_result(ref_id="ref-id-2"):
+    result = MagicMock()
+    result.ref.id = ref_id
+    return result
+
+
+# Platform-appropriate absolute base. tempfile.gettempdir() returns C:\... on
+# Windows and /tmp on POSIX, so containment via commonpath behaves naturally.
+_DEFAULT_BASE = os.path.join(__import__("tempfile").gettempdir(), "asset-enrichment-test-base")
+
+
+def _mocked_modules(*, enable_assets=True, register_file_in_place=None, directory=_DEFAULT_BASE):
+    return {
+        "comfy.cli_args": MagicMock(args=_make_args(enable_assets)),
+        "folder_paths": MagicMock(get_directory_by_type=MagicMock(return_value=directory)),
+        "app.assets.services.ingest": MagicMock(
+            register_file_in_place=register_file_in_place or MagicMock(return_value=_make_register_result()),
+            DependencyMissingError=type("DependencyMissingError", (Exception,), {}),
+        ),
+    }
+
+
+def _call(output_ui, *, enable_assets=True, file_exists=True, register_result=None, directory=_DEFAULT_BASE):
+    register_mock = MagicMock(return_value=register_result or _make_register_result())
+    mocked = _mocked_modules(
+        enable_assets=enable_assets,
+        register_file_in_place=register_mock,
+        directory=directory,
+    )
+
+    # Only os.path.isfile is patched — abspath/join must run natively so the
+    # containment check sees real platform paths.
+    with patch.dict("sys.modules", mocked), \
+         patch("os.path.isfile", return_value=file_exists):
+        import importlib
+        import comfy_execution.asset_enrichment as mod
+        importlib.reload(mod)
+        return mod.enrich_output_with_assets(output_ui)
+
+
+class TestEnrichOutputWithAssets(unittest.TestCase):
+
+    def test_disabled_returns_unchanged(self):
+        output = {"images": [{"filename": "a.png", "subfolder": "", "type": "output"}]}
+        result = _call(output, enable_assets=False)
+        self.assertNotIn("id", result["images"][0])
+
+    def test_non_list_value_passed_through(self):
+        output = {"text": "hello"}
+        result = _call(output)
+        self.assertEqual(result["text"], "hello")
+
+    def test_entry_without_filename_unchanged(self):
+        output = {"latent": [{"subfolder": "", "type": "output"}]}
+        result = _call(output)
+        self.assertNotIn("id", result["latent"][0])
+
+    def test_entry_without_type_unchanged(self):
+        output = {"data": [{"filename": "a.png", "subfolder": ""}]}
+        result = _call(output)
+        self.assertNotIn("id", result["data"][0])
+
+    def test_file_not_on_disk_unchanged(self):
+        output = {"images": [{"filename": "missing.png", "subfolder": "", "type": "output"}]}
+        result = _call(output, file_exists=False)
+        self.assertNotIn("id", result["images"][0])
+
+    def test_unknown_type_returns_none_directory_unchanged(self):
+        output = {"images": [{"filename": "a.png", "subfolder": "", "type": "unknown"}]}
+        result = _call(output, directory=None)
+        self.assertNotIn("id", result["images"][0])
+
+    def test_register_injects_only_id(self):
+        reg = _make_register_result(ref_id="inline-ref")
+        output = {"images": [{"filename": "new.png", "subfolder": "", "type": "output"}]}
+        result = _call(output, register_result=reg)
+        img = result["images"][0]
+        self.assertEqual(img["id"], "inline-ref")
+        # Only id is injected — no asset_hash, name, preview_url, size
+        self.assertNotIn("asset_hash", img)
+        self.assertNotIn("name", img)
+        self.assertNotIn("preview_url", img)
+        self.assertNotIn("size", img)
+
+    def test_register_called_per_entry(self):
+        register_mock = MagicMock(return_value=_make_register_result())
+        mocked = _mocked_modules(register_file_in_place=register_mock)
+        output = {
+            "images": [
+                {"filename": "a.png", "subfolder": "", "type": "output"},
+                {"filename": "b.png", "subfolder": "", "type": "output"},
+            ]
+        }
+
+        with patch.dict("sys.modules", mocked), \
+             patch("os.path.isfile", return_value=True):
+            import importlib
+            import comfy_execution.asset_enrichment as mod
+            importlib.reload(mod)
+            mod.enrich_output_with_assets(output)
+
+        self.assertEqual(register_mock.call_count, 2)
+
+    def test_original_entry_not_mutated(self):
+        orig = {"filename": "a.png", "subfolder": "", "type": "output"}
+        output = {"images": [orig]}
+        _call(output)
+        self.assertNotIn("id", orig)
+
+    def test_enrichment_error_does_not_block_sibling_entries(self):
+        call_count = [0]
+        good_reg = _make_register_result(ref_id="good-ref")
+
+        def register_side_effect(abs_path, name, tags):
+            call_count[0] += 1
+            if call_count[0] == 1:
+                raise RuntimeError("boom")
+            return good_reg
+
+        mocked = _mocked_modules(register_file_in_place=register_side_effect)
+
+        output = {
+            "images": [
+                {"filename": "bad.png", "subfolder": "", "type": "output"},
+                {"filename": "good.png", "subfolder": "", "type": "output"},
+            ]
+        }
+
+        with patch.dict("sys.modules", mocked), \
+             patch("os.path.isfile", return_value=True):
+            import importlib
+            import comfy_execution.asset_enrichment as mod
+            importlib.reload(mod)
+            result = mod.enrich_output_with_assets(output)
+
+        imgs = result["images"]
+        self.assertNotIn("id", imgs[0])
+        self.assertEqual(imgs[1]["id"], "good-ref")
+
+    def test_multiple_output_keys_all_enriched(self):
+        output = {
+            "images": [{"filename": "a.png", "subfolder": "", "type": "output"}],
+            "videos": [{"filename": "b.mp4", "subfolder": "", "type": "output"}],
+        }
+        result = _call(output)
+        self.assertIn("id", result["images"][0])
+        self.assertIn("id", result["videos"][0])
+
+    def test_none_entry_in_list_unchanged(self):
+        output = {"images": [None, {"filename": "a.png", "subfolder": "", "type": "output"}]}
+        result = _call(output)
+        self.assertIsNone(result["images"][0])
+        self.assertIn("id", result["images"][1])
+
+    def test_path_traversal_subfolder_skipped(self):
+        register_mock = MagicMock(return_value=_make_register_result())
+        mocked = _mocked_modules(register_file_in_place=register_mock)
+
+        output = {"images": [{"filename": "passwd", "subfolder": "../../etc", "type": "output"}]}
+
+        # Do NOT patch os.path.abspath — real resolution is required for the containment check.
+        with patch.dict("sys.modules", mocked), \
+             patch("os.path.isfile", return_value=True):
+            import importlib
+            import comfy_execution.asset_enrichment as mod
+            importlib.reload(mod)
+            result = mod.enrich_output_with_assets(output)
+
+        self.assertNotIn("id", result["images"][0])
+        register_mock.assert_not_called()
+
+    def test_absolute_filename_skipped(self):
+        register_mock = MagicMock(return_value=_make_register_result())
+        mocked = _mocked_modules(register_file_in_place=register_mock)
+
+        # Absolute filename — os.path.join discards earlier components when a later one is absolute.
+        absolute_filename = os.path.abspath(os.sep + "etc" + os.sep + "passwd")
+        output = {"images": [{"filename": absolute_filename, "subfolder": "", "type": "output"}]}
+
+        with patch.dict("sys.modules", mocked), \
+             patch("os.path.isfile", return_value=True):
+            import importlib
+            import comfy_execution.asset_enrichment as mod
+            importlib.reload(mod)
+            result = mod.enrich_output_with_assets(output)
+
+        self.assertNotIn("id", result["images"][0])
+        register_mock.assert_not_called()
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/execution/test_jobs.py

@@ -1,5 +1,7 @@
 """Unit tests for comfy_execution/jobs.py"""
 
+import pytest
+
 from comfy_execution.jobs import (
     JobStatus,
     is_previewable,
@@ -10,9 +12,50 @@ from comfy_execution.jobs import (
     get_outputs_summary,
     apply_sorting,
     has_3d_extension,
+    validate_job_id,
 )
 
 
+class TestValidateJobId:
+    """validate_job_id guards job creation: POST /prompt rejects ids it raises on."""
+
+    def test_canonical_form_passes_through(self):
+        cid = "a1b2c3d4-e5f6-7a89-b0c1-d2e3f4a5b6c7"
+        assert validate_job_id(cid) == cid
+
+    @pytest.mark.parametrize(
+        "variant",
+        [
+            "A1B2C3D4-E5F6-7A89-B0C1-D2E3F4A5B6C7",          # uppercase
+            "{a1b2c3d4-e5f6-7a89-b0c1-d2e3f4a5b6c7}",        # braced
+            "urn:uuid:a1b2c3d4-e5f6-7a89-b0c1-d2e3f4a5b6c7",  # URN
+            "a1b2c3d4e5f67a89b0c1d2e3f4a5b6c7",              # bare hex
+            " a1b2c3d4-e5f6-7a89-b0c1-d2e3f4a5b6c7 ",        # padded
+        ],
+    )
+    def test_non_canonical_spellings_rejected(self, variant):
+        # uuid.UUID parses all of these, but accepting them would silently
+        # rewrite the client's id (history keys, websocket events, and
+        # /interrupt matching all match the stored form exactly).
+        with pytest.raises(ValueError):
+            validate_job_id(variant)
+
+    @pytest.mark.parametrize(
+        "bad",
+        ["", "not-a-uuid", "prompt-123", "a1b2c3d4-e5f6-7a89-b0c1", "None"],
+    )
+    def test_non_uuid_strings_rejected(self, bad):
+        with pytest.raises(ValueError):
+            validate_job_id(bad)
+
+    @pytest.mark.parametrize("bad", [123, 1.5, True, None, ["a"], {"id": "x"}])
+    def test_non_strings_rejected(self, bad):
+        # uuid.UUID raises AttributeError/TypeError on non-strings; the helper
+        # must normalize those to ValueError so callers need one except clause.
+        with pytest.raises(ValueError):
+            validate_job_id(bad)
+
+
 class TestJobStatus:
     """Test JobStatus constants."""