Merge 7d0660f4e4 into fc0cb10bcb

Added max-parallel setting to enforce sequential execution in test jobs.

.github/workflows/test-ci.yml

@@ -20,6 +20,7 @@ jobs:
   test-stable:
     strategy:
       fail-fast: false
+      max-parallel: 1  # This forces sequential execution
       matrix:
         # os: [macos, linux, windows]
         # os: [macos, linux]
@@ -74,6 +75,7 @@ jobs:
   test-unix-nightly:
     strategy:
       fail-fast: false
+      max-parallel: 1  # This forces sequential execution
       matrix:
         # os: [macos, linux]
         os: [linux]

comfy/ops.py

@@ -427,12 +427,12 @@ def fp8_linear(self, input):
     input = torch.clamp(input, min=-448, max=448, out=input)
     input_fp8 = input.to(dtype).contiguous()
     layout_params_input = TensorCoreFP8Layout.Params(scale=scale_input, orig_dtype=input_dtype, orig_shape=tuple(input_fp8.shape))
-    quantized_input = QuantizedTensor(input_fp8, TensorCoreFP8Layout, layout_params_input)
+    quantized_input = QuantizedTensor(input_fp8, "TensorCoreFP8Layout", layout_params_input)
 
     # Wrap weight in QuantizedTensor - this enables unified dispatch
     # Call F.linear - __torch_dispatch__ routes to fp8_linear handler in quant_ops.py!
     layout_params_weight = TensorCoreFP8Layout.Params(scale=scale_weight, orig_dtype=input_dtype, orig_shape=tuple(w.shape))
-    quantized_weight = QuantizedTensor(w, TensorCoreFP8Layout, layout_params_weight)
+    quantized_weight = QuantizedTensor(w, "TensorCoreFP8Layout", layout_params_weight)
     o = torch.nn.functional.linear(quantized_input, quantized_weight, bias)
 
     uncast_bias_weight(self, w, bias, offload_stream)

comfy/quant_ops.py

@@ -13,6 +13,13 @@ try:
         get_layout_class,
     )
     _CK_AVAILABLE = True
+    if torch.version.cuda is None:
+        ck.registry.disable("cuda")
+    else:
+        cuda_version = tuple(map(int, str(torch.version.cuda).split('.')))
+        if cuda_version < (13,):
+            ck.registry.disable("cuda")
+
     ck.registry.disable("triton")
     for k, v in ck.list_backends().items():
         logging.info(f"Found comfy_kitchen backend {k}: {v}")

comfy_api/latest/_io.py

@@ -1113,6 +1113,22 @@ class DynamicSlot(ComfyTypeI):
             out_dict[input_type][finalized_id] = value
             out_dict["dynamic_paths"][finalized_id] = finalize_prefix(curr_prefix, curr_prefix[-1])
 
+@comfytype(io_type="IMAGECROP")
+class ImageCrop(ComfyTypeI):
+  """Widget for visual image cropping with x, y, width, height controls."""
+  Type = dict
+
+  class Input(WidgetInput):
+      def __init__(self, id: str, display_name: str=None, optional=False, tooltip: str=None,
+                   socketless: bool=True,
+                   default: dict=None):
+          super().__init__(id, display_name, optional, tooltip, None, default, socketless)
+          if default is None:
+              self.default = {"x": 0, "y": 0, "width": 512, "height": 512}
+
+      def as_dict(self):
+          return super().as_dict()
+
 DYNAMIC_INPUT_LOOKUP: dict[str, Callable[[dict[str, Any], dict[str, Any], tuple[str, dict[str, Any]], str, list[str] | None], None]] = {}
 def register_dynamic_input_func(io_type: str, func: Callable[[dict[str, Any], dict[str, Any], tuple[str, dict[str, Any]], str, list[str] | None], None]):
     DYNAMIC_INPUT_LOOKUP[io_type] = func
@@ -1958,4 +1974,5 @@ __all__ = [
     "add_to_dict_v1",
     "add_to_dict_v3",
     "V3Data",
+    "ImageCrop",
 ]

comfy_api_nodes/nodes_wan.py

@@ -13,7 +13,9 @@ from comfy_api_nodes.util import (
     poll_op,
     sync_op,
     tensor_to_base64_string,
+    upload_video_to_comfyapi,
     validate_audio_duration,
+    validate_video_duration,
 )
 
 
@@ -41,6 +43,12 @@ class Image2VideoInputField(BaseModel):
     audio_url: str | None = Field(None)
 
 
+class Reference2VideoInputField(BaseModel):
+    prompt: str = Field(...)
+    negative_prompt: str | None = Field(None)
+    reference_video_urls: list[str] = Field(...)
+
+
 class Txt2ImageParametersField(BaseModel):
     size: str = Field(...)
     n: int = Field(1, description="Number of images to generate.")  # we support only value=1
@@ -76,6 +84,14 @@ class Image2VideoParametersField(BaseModel):
     shot_type: str = Field("single")
 
 
+class Reference2VideoParametersField(BaseModel):
+    size: str = Field(...)
+    duration: int = Field(5, ge=5, le=15)
+    shot_type: str = Field("single")
+    seed: int = Field(..., ge=0, le=2147483647)
+    watermark: bool = Field(False)
+
+
 class Text2ImageTaskCreationRequest(BaseModel):
     model: str = Field(...)
     input: Text2ImageInputField = Field(...)
@@ -100,6 +116,12 @@ class Image2VideoTaskCreationRequest(BaseModel):
     parameters: Image2VideoParametersField = Field(...)
 
 
+class Reference2VideoTaskCreationRequest(BaseModel):
+    model: str = Field(...)
+    input: Reference2VideoInputField = Field(...)
+    parameters: Reference2VideoParametersField = Field(...)
+
+
 class TaskCreationOutputField(BaseModel):
     task_id: str = Field(...)
     task_status: str = Field(...)
@@ -721,6 +743,143 @@ class WanImageToVideoApi(IO.ComfyNode):
         return IO.NodeOutput(await download_url_to_video_output(response.output.video_url))
 
 
+class WanReferenceVideoApi(IO.ComfyNode):
+    @classmethod
+    def define_schema(cls):
+        return IO.Schema(
+            node_id="WanReferenceVideoApi",
+            display_name="Wan Reference to Video",
+            category="api node/video/Wan",
+            description="Use the character and voice from input videos, combined with a prompt, "
+            "to generate a new video that maintains character consistency.",
+            inputs=[
+                IO.Combo.Input("model", options=["wan2.6-r2v"]),
+                IO.String.Input(
+                    "prompt",
+                    multiline=True,
+                    default="",
+                    tooltip="Prompt describing the elements and visual features. Supports English and Chinese. "
+                    "Use identifiers such as `character1` and `character2` to refer to the reference characters.",
+                ),
+                IO.String.Input(
+                    "negative_prompt",
+                    multiline=True,
+                    default="",
+                    tooltip="Negative prompt describing what to avoid.",
+                ),
+                IO.Autogrow.Input(
+                    "reference_videos",
+                    template=IO.Autogrow.TemplateNames(
+                        IO.Video.Input("reference_video"),
+                        names=["character1", "character2", "character3"],
+                        min=1,
+                    ),
+                ),
+                IO.Combo.Input(
+                    "size",
+                    options=[
+                        "720p: 1:1 (960x960)",
+                        "720p: 16:9 (1280x720)",
+                        "720p: 9:16 (720x1280)",
+                        "720p: 4:3 (1088x832)",
+                        "720p: 3:4 (832x1088)",
+                        "1080p: 1:1 (1440x1440)",
+                        "1080p: 16:9 (1920x1080)",
+                        "1080p: 9:16 (1080x1920)",
+                        "1080p: 4:3 (1632x1248)",
+                        "1080p: 3:4 (1248x1632)",
+                    ],
+                ),
+                IO.Int.Input(
+                    "duration",
+                    default=5,
+                    min=5,
+                    max=10,
+                    step=5,
+                    display_mode=IO.NumberDisplay.slider,
+                ),
+                IO.Int.Input(
+                    "seed",
+                    default=0,
+                    min=0,
+                    max=2147483647,
+                    step=1,
+                    display_mode=IO.NumberDisplay.number,
+                    control_after_generate=True,
+                ),
+                IO.Combo.Input(
+                    "shot_type",
+                    options=["single", "multi"],
+                    tooltip="Specifies the shot type for the generated video, that is, whether the video is a "
+                    "single continuous shot or multiple shots with cuts.",
+                ),
+                IO.Boolean.Input(
+                    "watermark",
+                    default=False,
+                    tooltip="Whether to add an AI-generated watermark to the result.",
+                ),
+            ],
+            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,
+        )
+
+    @classmethod
+    async def execute(
+        cls,
+        model: str,
+        prompt: str,
+        negative_prompt: str,
+        reference_videos: IO.Autogrow.Type,
+        size: str,
+        duration: int,
+        seed: int,
+        shot_type: str,
+        watermark: bool,
+    ):
+        reference_video_urls = []
+        for i in reference_videos:
+            validate_video_duration(reference_videos[i], min_duration=2, max_duration=30)
+        for i in reference_videos:
+            reference_video_urls.append(await upload_video_to_comfyapi(cls, reference_videos[i]))
+        width, height = RES_IN_PARENS.search(size).groups()
+        initial_response = await sync_op(
+            cls,
+            ApiEndpoint(path="/proxy/wan/api/v1/services/aigc/video-generation/video-synthesis", method="POST"),
+            response_model=TaskCreationResponse,
+            data=Reference2VideoTaskCreationRequest(
+                model=model,
+                input=Reference2VideoInputField(
+                    prompt=prompt, negative_prompt=negative_prompt, reference_video_urls=reference_video_urls
+                ),
+                parameters=Reference2VideoParametersField(
+                    size=f"{width}*{height}",
+                    duration=duration,
+                    shot_type=shot_type,
+                    watermark=watermark,
+                    seed=seed,
+                ),
+            ),
+        )
+        if not initial_response.output:
+            raise Exception(f"An unknown error occurred: {initial_response.code} - {initial_response.message}")
+        response = await poll_op(
+            cls,
+            ApiEndpoint(path=f"/proxy/wan/api/v1/tasks/{initial_response.output.task_id}"),
+            response_model=VideoTaskStatusResponse,
+            status_extractor=lambda x: x.output.task_status,
+            poll_interval=6,
+            max_poll_attempts=280,
+        )
+        return IO.NodeOutput(await download_url_to_video_output(response.output.video_url))
+
+
 class WanApiExtension(ComfyExtension):
     @override
     async def get_node_list(self) -> list[type[IO.ComfyNode]]:
@@ -729,6 +888,7 @@ class WanApiExtension(ComfyExtension):
             WanImageToImageApi,
             WanTextToVideoApi,
             WanImageToVideoApi,
+            WanReferenceVideoApi,
         ]
 
 

comfy_api_nodes/util/upload_helpers.py

@@ -119,7 +119,7 @@ async def upload_video_to_comfyapi(
             raise ValueError(f"Could not verify video duration from source: {e}") from e
 
     upload_mime_type = f"video/{container.value.lower()}"
-    filename = f"uploaded_video.{container.value.lower()}"
+    filename = f"{uuid.uuid4()}.{container.value.lower()}"
 
     # Convert VideoInput to BytesIO using specified container/codec
     video_bytes_io = BytesIO()

comfy_extras/nodes_images.py

@@ -26,16 +26,18 @@ class ImageCrop(IO.ComfyNode):
             category="image/transform",
             inputs=[
                 IO.Image.Input("image"),
-                IO.Int.Input("width", default=512, min=1, max=nodes.MAX_RESOLUTION, step=1),
-                IO.Int.Input("height", default=512, min=1, max=nodes.MAX_RESOLUTION, step=1),
-                IO.Int.Input("x", default=0, min=0, max=nodes.MAX_RESOLUTION, step=1),
-                IO.Int.Input("y", default=0, min=0, max=nodes.MAX_RESOLUTION, step=1),
+                IO.ImageCrop.Input("crop_region"),
             ],
             outputs=[IO.Image.Output()],
         )
 
     @classmethod
-    def execute(cls, image, width, height, x, y) -> IO.NodeOutput:
+    def execute(cls, image, crop_region) -> IO.NodeOutput:
+        x = crop_region.get("x", 0)
+        y = crop_region.get("y", 0)
+        width = crop_region.get("width", 512)
+        height = crop_region.get("height", 512)
+
         x = min(x, image.shape[2] - 1)
         y = min(y, image.shape[1] - 1)
         to_x = width + x

comfyui_version.py

@@ -1,3 +1,3 @@
 # This file is automatically generated by the build process when version is
 # updated in pyproject.toml.
-__version__ = "0.7.0"
+__version__ = "0.8.0"

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "ComfyUI"
-version = "0.7.0"
+version = "0.8.0"
 readme = "README.md"
 license = { file = "LICENSE" }
 requires-python = ">=3.10"

requirements.txt

@@ -21,7 +21,7 @@ psutil
 alembic
 SQLAlchemy
 av>=14.2.0
-comfy-kitchen>=0.2.2
+comfy-kitchen>=0.2.3
 
 #non essential dependencies:
 kornia>=0.7.1