Merge branch 'master' into automation/comfyui-frontend-bump

comfy/ldm/anima/model.py

@@ -179,8 +179,8 @@ class LLMAdapter(nn.Module):
             if source_attention_mask.ndim == 2:
                 source_attention_mask = source_attention_mask.unsqueeze(1).unsqueeze(1)
 
-        x = self.in_proj(self.embed(target_input_ids))
         context = source_hidden_states
+        x = self.in_proj(self.embed(target_input_ids, out_dtype=context.dtype))
         position_ids = torch.arange(x.shape[1], device=x.device).unsqueeze(0)
         position_ids_context = torch.arange(context.shape[1], device=x.device).unsqueeze(0)
         position_embeddings = self.rotary_emb(x, position_ids)

comfy/ops.py

@@ -79,7 +79,7 @@ def cast_to_input(weight, input, non_blocking=False, copy=True):
     return comfy.model_management.cast_to(weight, input.dtype, input.device, non_blocking=non_blocking, copy=copy)
 
 
-def cast_bias_weight_with_vbar(s, dtype, device, bias_dtype, non_blocking, compute_dtype):
+def cast_bias_weight_with_vbar(s, dtype, device, bias_dtype, non_blocking, compute_dtype, want_requant):
     offload_stream = None
     xfer_dest = None
 
@@ -170,10 +170,10 @@ def cast_bias_weight_with_vbar(s, dtype, device, bias_dtype, non_blocking, compu
             #FIXME: this is not accurate, we need to be sensitive to the compute dtype
             x = lowvram_fn(x)
             if (isinstance(orig, QuantizedTensor) and
-                (orig.dtype == dtype and len(fns) == 0 or update_weight)):
+                (want_requant and len(fns) == 0 or update_weight)):
                 seed = comfy.utils.string_to_seed(s.seed_key)
                 y = QuantizedTensor.from_float(x, s.layout_type, scale="recalculate", stochastic_rounding=seed)
-                if orig.dtype == dtype and len(fns) == 0:
+                if want_requant and len(fns) == 0:
                     #The layer actually wants our freshly saved QT
                     x = y
             elif update_weight:
@@ -194,7 +194,7 @@ def cast_bias_weight_with_vbar(s, dtype, device, bias_dtype, non_blocking, compu
     return weight, bias, (offload_stream, device if signature is not None else None, None)
 
 
-def cast_bias_weight(s, input=None, dtype=None, device=None, bias_dtype=None, offloadable=False, compute_dtype=None):
+def cast_bias_weight(s, input=None, dtype=None, device=None, bias_dtype=None, offloadable=False, compute_dtype=None, want_requant=False):
     # NOTE: offloadable=False is a a legacy and if you are a custom node author reading this please pass
     # offloadable=True and call uncast_bias_weight() after your last usage of the weight/bias. This
     # will add async-offload support to your cast and improve performance.
@@ -212,7 +212,7 @@ 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"):
-        return cast_bias_weight_with_vbar(s, dtype, device, bias_dtype, non_blocking, compute_dtype)
+        return cast_bias_weight_with_vbar(s, dtype, device, bias_dtype, non_blocking, compute_dtype, want_requant)
 
     if offloadable and (device != s.weight.device or
                         (s.bias is not None and device != s.bias.device)):
@@ -850,8 +850,8 @@ def mixed_precision_ops(quant_config={}, compute_dtype=torch.bfloat16, full_prec
             def _forward(self, input, weight, bias):
                 return torch.nn.functional.linear(input, weight, bias)
 
-            def forward_comfy_cast_weights(self, input, compute_dtype=None):
+            def forward_comfy_cast_weights(self, input, compute_dtype=None, want_requant=False):
-                weight, bias, offload_stream = cast_bias_weight(self, input, offloadable=True, compute_dtype=compute_dtype)
+                weight, bias, offload_stream = cast_bias_weight(self, input, offloadable=True, compute_dtype=compute_dtype, want_requant=want_requant)
                 x = self._forward(input, weight, bias)
                 uncast_bias_weight(self, weight, bias, offload_stream)
                 return x
@@ -881,8 +881,7 @@ def mixed_precision_ops(quant_config={}, compute_dtype=torch.bfloat16, full_prec
                             scale = comfy.model_management.cast_to_device(scale, input.device, None)
                         input = QuantizedTensor.from_float(input_reshaped, self.layout_type, scale=scale)
 
-
+                output = self.forward_comfy_cast_weights(input, compute_dtype, want_requant=isinstance(input, QuantizedTensor))
-                output = self.forward_comfy_cast_weights(input, compute_dtype)
 
                 # Reshape output back to 3D if input was 3D
                 if reshaped_3d:

comfy_api_nodes/apis/recraft.py

@@ -198,11 +198,6 @@ dict_recraft_substyles_v3 = {
 }
 
 
-class RecraftModel(str, Enum):
-    recraftv3 = 'recraftv3'
-    recraftv2 = 'recraftv2'
-
-
 class RecraftImageSize(str, Enum):
     res_1024x1024 = '1024x1024'
     res_1365x1024 = '1365x1024'
@@ -221,6 +216,41 @@ class RecraftImageSize(str, Enum):
     res_1707x1024 = '1707x1024'
 
 
+RECRAFT_V4_SIZES = [
+    "1024x1024",
+    "1536x768",
+    "768x1536",
+    "1280x832",
+    "832x1280",
+    "1216x896",
+    "896x1216",
+    "1152x896",
+    "896x1152",
+    "832x1344",
+    "1280x896",
+    "896x1280",
+    "1344x768",
+    "768x1344",
+]
+
+RECRAFT_V4_PRO_SIZES = [
+    "2048x2048",
+    "3072x1536",
+    "1536x3072",
+    "2560x1664",
+    "1664x2560",
+    "2432x1792",
+    "1792x2432",
+    "2304x1792",
+    "1792x2304",
+    "1664x2688",
+    "1434x1024",
+    "1024x1434",
+    "2560x1792",
+    "1792x2560",
+]
+
+
 class RecraftColorObject(BaseModel):
     rgb: list[int] = Field(..., description='An array of 3 integer values in range of 0...255 defining RGB Color Model')
 
@@ -234,17 +264,16 @@ class RecraftControlsObject(BaseModel):
 
 class RecraftImageGenerationRequest(BaseModel):
     prompt: str = Field(..., description='The text prompt describing the image to generate')
-    size: RecraftImageSize | None = Field(None, description='The size of the generated image (e.g., "1024x1024")')
+    size: str | None = Field(None, description='The size of the generated image (e.g., "1024x1024")')
     n: int = Field(..., description='The number of images to generate')
     negative_prompt: str | None = Field(None, description='A text description of undesired elements on an image')
-    model: RecraftModel | None = Field(RecraftModel.recraftv3, description='The model to use for generation (e.g., "recraftv3")')
+    model: str = Field(...)
     style: str | None = Field(None, description='The style to apply to the generated image (e.g., "digital_illustration")')
     substyle: str | None = Field(None, description='The substyle to apply to the generated image, depending on the style input')
     controls: RecraftControlsObject | None = Field(None, description='A set of custom parameters to tweak generation process')
     style_id: str | None = Field(None, description='Use a previously uploaded style as a reference; UUID')
     strength: float | None = Field(None, description='Defines the difference with the original image, should lie in [0, 1], where 0 means almost identical, and 1 means miserable similarity')
     random_seed: int | None = Field(None, description="Seed for video generation")
-    # text_layout
 
 
 class RecraftReturnedObject(BaseModel):

comfy_api_nodes/nodes_recraft.py

@@ -1,5 +1,4 @@
 from io import BytesIO
-from typing import Optional, Union
 
 import aiohttp
 import torch
@@ -9,6 +8,8 @@ from typing_extensions import override
 from comfy.utils import ProgressBar
 from comfy_api.latest import IO, ComfyExtension
 from comfy_api_nodes.apis.recraft import (
+    RECRAFT_V4_PRO_SIZES,
+    RECRAFT_V4_SIZES,
     RecraftColor,
     RecraftColorChain,
     RecraftControls,
@@ -18,7 +19,6 @@ from comfy_api_nodes.apis.recraft import (
     RecraftImageGenerationResponse,
     RecraftImageSize,
     RecraftIO,
-    RecraftModel,
     RecraftStyle,
     RecraftStyleV3,
     get_v3_substyles,
@@ -39,7 +39,7 @@ async def handle_recraft_file_request(
     cls: type[IO.ComfyNode],
     image: torch.Tensor,
     path: str,
-    mask: Optional[torch.Tensor] = None,
+    mask: torch.Tensor | None = None,
     total_pixels: int = 4096 * 4096,
     timeout: int = 1024,
     request=None,
@@ -73,11 +73,11 @@ async def handle_recraft_file_request(
 def recraft_multipart_parser(
     data,
     parent_key=None,
-    formatter: Optional[type[callable]] = None,
+    formatter: type[callable] | None = None,
-    converted_to_check: Optional[list[list]] = None,
+    converted_to_check: list[list] | None = None,
     is_list: bool = False,
     return_mode: str = "formdata",  # "dict" | "formdata"
-) -> Union[dict, aiohttp.FormData]:
+) -> dict | aiohttp.FormData:
     """
     Formats data such that multipart/form-data will work with aiohttp library when both files and data are present.
 
@@ -309,7 +309,7 @@ class RecraftStyleInfiniteStyleLibrary(IO.ComfyNode):
             node_id="RecraftStyleV3InfiniteStyleLibrary",
             display_name="Recraft Style - Infinite Style Library",
             category="api node/image/Recraft",
-            description="Select style based on preexisting UUID from Recraft's Infinite Style Library.",
+            description="Choose style based on preexisting UUID from Recraft's Infinite Style Library.",
             inputs=[
                 IO.String.Input("style_id", default="", tooltip="UUID of style from Infinite Style Library."),
             ],
@@ -485,7 +485,7 @@ class RecraftTextToImageNode(IO.ComfyNode):
             data=RecraftImageGenerationRequest(
                 prompt=prompt,
                 negative_prompt=negative_prompt,
-                model=RecraftModel.recraftv3,
+                model="recraftv3",
                 size=size,
                 n=n,
                 style=recraft_style.style,
@@ -598,7 +598,7 @@ class RecraftImageToImageNode(IO.ComfyNode):
         request = RecraftImageGenerationRequest(
             prompt=prompt,
             negative_prompt=negative_prompt,
-            model=RecraftModel.recraftv3,
+            model="recraftv3",
             n=n,
             strength=round(strength, 2),
             style=recraft_style.style,
@@ -698,7 +698,7 @@ class RecraftImageInpaintingNode(IO.ComfyNode):
         request = RecraftImageGenerationRequest(
             prompt=prompt,
             negative_prompt=negative_prompt,
-            model=RecraftModel.recraftv3,
+            model="recraftv3",
             n=n,
             style=recraft_style.style,
             substyle=recraft_style.substyle,
@@ -810,7 +810,7 @@ class RecraftTextToVectorNode(IO.ComfyNode):
             data=RecraftImageGenerationRequest(
                 prompt=prompt,
                 negative_prompt=negative_prompt,
-                model=RecraftModel.recraftv3,
+                model="recraftv3",
                 size=size,
                 n=n,
                 style=recraft_style.style,
@@ -933,7 +933,7 @@ class RecraftReplaceBackgroundNode(IO.ComfyNode):
         request = RecraftImageGenerationRequest(
             prompt=prompt,
             negative_prompt=negative_prompt,
-            model=RecraftModel.recraftv3,
+            model="recraftv3",
             n=n,
             style=recraft_style.style,
             substyle=recraft_style.substyle,
@@ -1078,6 +1078,252 @@ class RecraftCreativeUpscaleNode(RecraftCrispUpscaleNode):
         )
 
 
+class RecraftV4TextToImageNode(IO.ComfyNode):
+    @classmethod
+    def define_schema(cls):
+        return IO.Schema(
+            node_id="RecraftV4TextToImageNode",
+            display_name="Recraft V4 Text to Image",
+            category="api node/image/Recraft",
+            description="Generates images using Recraft V4 or V4 Pro models.",
+            inputs=[
+                IO.String.Input(
+                    "prompt",
+                    multiline=True,
+                    tooltip="Prompt for the image generation. Maximum 10,000 characters.",
+                ),
+                IO.String.Input(
+                    "negative_prompt",
+                    multiline=True,
+                    tooltip="An optional text description of undesired elements on an image.",
+                ),
+                IO.DynamicCombo.Input(
+                    "model",
+                    options=[
+                        IO.DynamicCombo.Option(
+                            "recraftv4",
+                            [
+                                IO.Combo.Input(
+                                    "size",
+                                    options=RECRAFT_V4_SIZES,
+                                    default="1024x1024",
+                                    tooltip="The size of the generated image.",
+                                ),
+                            ],
+                        ),
+                        IO.DynamicCombo.Option(
+                            "recraftv4_pro",
+                            [
+                                IO.Combo.Input(
+                                    "size",
+                                    options=RECRAFT_V4_PRO_SIZES,
+                                    default="2048x2048",
+                                    tooltip="The size of the generated image.",
+                                ),
+                            ],
+                        ),
+                    ],
+                    tooltip="The model to use for generation.",
+                ),
+                IO.Int.Input(
+                    "n",
+                    default=1,
+                    min=1,
+                    max=6,
+                    tooltip="The number of images to generate.",
+                ),
+                IO.Int.Input(
+                    "seed",
+                    default=0,
+                    min=0,
+                    max=0xFFFFFFFFFFFFFFFF,
+                    control_after_generate=True,
+                    tooltip="Seed to determine if node should re-run; "
+                    "actual results are nondeterministic regardless of seed.",
+                ),
+                IO.Custom(RecraftIO.CONTROLS).Input(
+                    "recraft_controls",
+                    tooltip="Optional additional controls over the generation via the Recraft Controls node.",
+                    optional=True,
+                ),
+            ],
+            outputs=[
+                IO.Image.Output(),
+            ],
+            hidden=[
+                IO.Hidden.auth_token_comfy_org,
+                IO.Hidden.api_key_comfy_org,
+                IO.Hidden.unique_id,
+            ],
+            is_api_node=True,
+            price_badge=IO.PriceBadge(
+                depends_on=IO.PriceBadgeDepends(widgets=["model", "n"]),
+                expr="""
+                (
+                    $prices := {"recraftv4": 0.04, "recraftv4_pro": 0.25};
+                    {"type":"usd","usd": $lookup($prices, widgets.model) * widgets.n}
+                )
+                """,
+            ),
+        )
+
+    @classmethod
+    async def execute(
+        cls,
+        prompt: str,
+        negative_prompt: str,
+        model: dict,
+        n: int,
+        seed: int,
+        recraft_controls: RecraftControls | None = None,
+    ) -> IO.NodeOutput:
+        validate_string(prompt, strip_whitespace=False, min_length=1, max_length=10000)
+        response = await sync_op(
+            cls,
+            ApiEndpoint(path="/proxy/recraft/image_generation", method="POST"),
+            response_model=RecraftImageGenerationResponse,
+            data=RecraftImageGenerationRequest(
+                prompt=prompt,
+                negative_prompt=negative_prompt if negative_prompt else None,
+                model=model["model"],
+                size=model["size"],
+                n=n,
+                controls=recraft_controls.create_api_model() if recraft_controls else None,
+            ),
+            max_retries=1,
+        )
+        images = []
+        for data in response.data:
+            with handle_recraft_image_output():
+                image = bytesio_to_image_tensor(await download_url_as_bytesio(data.url, timeout=1024))
+            if len(image.shape) < 4:
+                image = image.unsqueeze(0)
+            images.append(image)
+        return IO.NodeOutput(torch.cat(images, dim=0))
+
+
+class RecraftV4TextToVectorNode(IO.ComfyNode):
+    @classmethod
+    def define_schema(cls):
+        return IO.Schema(
+            node_id="RecraftV4TextToVectorNode",
+            display_name="Recraft V4 Text to Vector",
+            category="api node/image/Recraft",
+            description="Generates SVG using Recraft V4 or V4 Pro models.",
+            inputs=[
+                IO.String.Input(
+                    "prompt",
+                    multiline=True,
+                    tooltip="Prompt for the image generation. Maximum 10,000 characters.",
+                ),
+                IO.String.Input(
+                    "negative_prompt",
+                    multiline=True,
+                    tooltip="An optional text description of undesired elements on an image.",
+                ),
+                IO.DynamicCombo.Input(
+                    "model",
+                    options=[
+                        IO.DynamicCombo.Option(
+                            "recraftv4",
+                            [
+                                IO.Combo.Input(
+                                    "size",
+                                    options=RECRAFT_V4_SIZES,
+                                    default="1024x1024",
+                                    tooltip="The size of the generated image.",
+                                ),
+                            ],
+                        ),
+                        IO.DynamicCombo.Option(
+                            "recraftv4_pro",
+                            [
+                                IO.Combo.Input(
+                                    "size",
+                                    options=RECRAFT_V4_PRO_SIZES,
+                                    default="2048x2048",
+                                    tooltip="The size of the generated image.",
+                                ),
+                            ],
+                        ),
+                    ],
+                    tooltip="The model to use for generation.",
+                ),
+                IO.Int.Input(
+                    "n",
+                    default=1,
+                    min=1,
+                    max=6,
+                    tooltip="The number of images to generate.",
+                ),
+                IO.Int.Input(
+                    "seed",
+                    default=0,
+                    min=0,
+                    max=0xFFFFFFFFFFFFFFFF,
+                    control_after_generate=True,
+                    tooltip="Seed to determine if node should re-run; "
+                    "actual results are nondeterministic regardless of seed.",
+                ),
+                IO.Custom(RecraftIO.CONTROLS).Input(
+                    "recraft_controls",
+                    tooltip="Optional additional controls over the generation via the Recraft Controls node.",
+                    optional=True,
+                ),
+            ],
+            outputs=[
+                IO.SVG.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(
+                depends_on=IO.PriceBadgeDepends(widgets=["model", "n"]),
+                expr="""
+                (
+                    $prices := {"recraftv4": 0.08, "recraftv4_pro": 0.30};
+                    {"type":"usd","usd": $lookup($prices, widgets.model) * widgets.n}
+                )
+                """,
+            ),
+        )
+
+    @classmethod
+    async def execute(
+        cls,
+        prompt: str,
+        negative_prompt: str,
+        model: dict,
+        n: int,
+        seed: int,
+        recraft_controls: RecraftControls | None = None,
+    ) -> IO.NodeOutput:
+        validate_string(prompt, strip_whitespace=False, min_length=1, max_length=10000)
+        response = await sync_op(
+            cls,
+            ApiEndpoint(path="/proxy/recraft/image_generation", method="POST"),
+            response_model=RecraftImageGenerationResponse,
+            data=RecraftImageGenerationRequest(
+                prompt=prompt,
+                negative_prompt=negative_prompt if negative_prompt else None,
+                model=model["model"],
+                size=model["size"],
+                n=n,
+                style="vector_illustration",
+                substyle=None,
+                controls=recraft_controls.create_api_model() if recraft_controls else None,
+            ),
+            max_retries=1,
+        )
+        svg_data = []
+        for data in response.data:
+            svg_data.append(await download_url_as_bytesio(data.url, timeout=1024))
+        return IO.NodeOutput(SVG(svg_data))
+
+
 class RecraftExtension(ComfyExtension):
     @override
     async def get_node_list(self) -> list[type[IO.ComfyNode]]:
@@ -1098,6 +1344,8 @@ class RecraftExtension(ComfyExtension):
             RecraftCreateStyleNode,
             RecraftColorRGBNode,
             RecraftControlsNode,
+            RecraftV4TextToImageNode,
+            RecraftV4TextToVectorNode,
         ]
 
 

comfy_api_nodes/nodes_vidu.py

@@ -54,6 +54,7 @@ async def execute_task(
         response_model=TaskStatusResponse,
         status_extractor=lambda r: r.state,
         progress_extractor=lambda r: r.progress,
+        price_extractor=lambda r: r.credits * 0.005 if r.credits is not None else None,
         max_poll_attempts=max_poll_attempts,
     )
     if not response.creations:
@@ -1306,6 +1307,36 @@ class Vidu3TextToVideoNode(IO.ComfyNode):
                                 ),
                             ],
                         ),
+                        IO.DynamicCombo.Option(
+                            "viduq3-turbo",
+                            [
+                                IO.Combo.Input(
+                                    "aspect_ratio",
+                                    options=["16:9", "9:16", "3:4", "4:3", "1:1"],
+                                    tooltip="The aspect ratio of the output video.",
+                                ),
+                                IO.Combo.Input(
+                                    "resolution",
+                                    options=["720p", "1080p"],
+                                    tooltip="Resolution of the output video.",
+                                ),
+                                IO.Int.Input(
+                                    "duration",
+                                    default=5,
+                                    min=1,
+                                    max=16,
+                                    step=1,
+                                    display_mode=IO.NumberDisplay.slider,
+                                    tooltip="Duration of the output video in seconds.",
+                                ),
+                                IO.Boolean.Input(
+                                    "audio",
+                                    default=False,
+                                    tooltip="When enabled, outputs video with sound "
+                                    "(including dialogue and sound effects).",
+                                ),
+                            ],
+                        ),
                     ],
                     tooltip="Model to use for video generation.",
                 ),
@@ -1334,13 +1365,20 @@ class Vidu3TextToVideoNode(IO.ComfyNode):
             ],
             is_api_node=True,
             price_badge=IO.PriceBadge(
-                depends_on=IO.PriceBadgeDepends(widgets=["model.duration", "model.resolution"]),
+                depends_on=IO.PriceBadgeDepends(widgets=["model", "model.duration", "model.resolution"]),
                 expr="""
                 (
                   $res := $lookup(widgets, "model.resolution");
-                  $base := $lookup({"720p": 0.075, "1080p": 0.1}, $res);
+                  $d := $lookup(widgets, "model.duration");
-                  $perSec := $lookup({"720p": 0.025, "1080p": 0.05}, $res);
+                  $contains(widgets.model, "turbo")
-                  {"type":"usd","usd": $base + $perSec * ($lookup(widgets, "model.duration") - 1)}
+                    ? (
+                        $rate := $lookup({"720p": 0.06, "1080p": 0.08}, $res);
+                        {"type":"usd","usd": $rate * $d}
+                      )
+                    : (
+                        $rate := $lookup({"720p": 0.15, "1080p": 0.16}, $res);
+                        {"type":"usd","usd": $rate * $d}
+                      )
                 )
                 """,
             ),
@@ -1409,6 +1447,31 @@ class Vidu3ImageToVideoNode(IO.ComfyNode):
                                 ),
                             ],
                         ),
+                        IO.DynamicCombo.Option(
+                            "viduq3-turbo",
+                            [
+                                IO.Combo.Input(
+                                    "resolution",
+                                    options=["720p", "1080p"],
+                                    tooltip="Resolution of the output video.",
+                                ),
+                                IO.Int.Input(
+                                    "duration",
+                                    default=5,
+                                    min=1,
+                                    max=16,
+                                    step=1,
+                                    display_mode=IO.NumberDisplay.slider,
+                                    tooltip="Duration of the output video in seconds.",
+                                ),
+                                IO.Boolean.Input(
+                                    "audio",
+                                    default=False,
+                                    tooltip="When enabled, outputs video with sound "
+                                    "(including dialogue and sound effects).",
+                                ),
+                            ],
+                        ),
                     ],
                     tooltip="Model to use for video generation.",
                 ),
@@ -1442,13 +1505,20 @@ class Vidu3ImageToVideoNode(IO.ComfyNode):
             ],
             is_api_node=True,
             price_badge=IO.PriceBadge(
-                depends_on=IO.PriceBadgeDepends(widgets=["model.duration", "model.resolution"]),
+                depends_on=IO.PriceBadgeDepends(widgets=["model", "model.duration", "model.resolution"]),
                 expr="""
                 (
                   $res := $lookup(widgets, "model.resolution");
-                  $base := $lookup({"720p": 0.075, "1080p": 0.275, "2k": 0.35}, $res);
+                  $d := $lookup(widgets, "model.duration");
-                  $perSec := $lookup({"720p": 0.05, "1080p": 0.075, "2k": 0.075}, $res);
+                  $contains(widgets.model, "turbo")
-                  {"type":"usd","usd": $base + $perSec * ($lookup(widgets, "model.duration") - 1)}
+                    ? (
+                        $rate := $lookup({"720p": 0.06, "1080p": 0.08}, $res);
+                        {"type":"usd","usd": $rate * $d}
+                      )
+                    : (
+                        $rate := $lookup({"720p": 0.15, "1080p": 0.16, "2k": 0.2}, $res);
+                        {"type":"usd","usd": $rate * $d}
+                      )
                 )
                 """,
             ),
@@ -1481,6 +1551,145 @@ class Vidu3ImageToVideoNode(IO.ComfyNode):
         return IO.NodeOutput(await download_url_to_video_output(results[0].url))
 
 
+class Vidu3StartEndToVideoNode(IO.ComfyNode):
+
+    @classmethod
+    def define_schema(cls):
+        return IO.Schema(
+            node_id="Vidu3StartEndToVideoNode",
+            display_name="Vidu Q3 Start/End Frame-to-Video Generation",
+            category="api node/video/Vidu",
+            description="Generate a video from a start frame, an end frame, and a prompt.",
+            inputs=[
+                IO.DynamicCombo.Input(
+                    "model",
+                    options=[
+                        IO.DynamicCombo.Option(
+                            "viduq3-pro",
+                            [
+                                IO.Combo.Input(
+                                    "resolution",
+                                    options=["720p", "1080p"],
+                                    tooltip="Resolution of the output video.",
+                                ),
+                                IO.Int.Input(
+                                    "duration",
+                                    default=5,
+                                    min=1,
+                                    max=16,
+                                    step=1,
+                                    display_mode=IO.NumberDisplay.slider,
+                                    tooltip="Duration of the output video in seconds.",
+                                ),
+                                IO.Boolean.Input(
+                                    "audio",
+                                    default=False,
+                                    tooltip="When enabled, outputs video with sound "
+                                    "(including dialogue and sound effects).",
+                                ),
+                            ],
+                        ),
+                        IO.DynamicCombo.Option(
+                            "viduq3-turbo",
+                            [
+                                IO.Combo.Input(
+                                    "resolution",
+                                    options=["720p", "1080p"],
+                                    tooltip="Resolution of the output video.",
+                                ),
+                                IO.Int.Input(
+                                    "duration",
+                                    default=5,
+                                    min=1,
+                                    max=16,
+                                    step=1,
+                                    display_mode=IO.NumberDisplay.slider,
+                                    tooltip="Duration of the output video in seconds.",
+                                ),
+                                IO.Boolean.Input(
+                                    "audio",
+                                    default=False,
+                                    tooltip="When enabled, outputs video with sound "
+                                    "(including dialogue and sound effects).",
+                                ),
+                            ],
+                        ),
+                    ],
+                    tooltip="Model to use for video generation.",
+                ),
+                IO.Image.Input("first_frame"),
+                IO.Image.Input("end_frame"),
+                IO.String.Input(
+                    "prompt",
+                    multiline=True,
+                    tooltip="Prompt description (max 2000 characters).",
+                ),
+                IO.Int.Input(
+                    "seed",
+                    default=1,
+                    min=0,
+                    max=2147483647,
+                    step=1,
+                    display_mode=IO.NumberDisplay.number,
+                    control_after_generate=True,
+                ),
+            ],
+            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(
+                depends_on=IO.PriceBadgeDepends(widgets=["model", "model.duration", "model.resolution"]),
+                expr="""
+                (
+                  $res := $lookup(widgets, "model.resolution");
+                  $d := $lookup(widgets, "model.duration");
+                  $contains(widgets.model, "turbo")
+                    ? (
+                        $rate := $lookup({"720p": 0.06, "1080p": 0.08}, $res);
+                        {"type":"usd","usd": $rate * $d}
+                      )
+                    : (
+                        $rate := $lookup({"720p": 0.15, "1080p": 0.16}, $res);
+                        {"type":"usd","usd": $rate * $d}
+                      )
+                )
+                """,
+            ),
+        )
+
+    @classmethod
+    async def execute(
+        cls,
+        model: dict,
+        first_frame: Input.Image,
+        end_frame: Input.Image,
+        prompt: str,
+        seed: int,
+    ) -> IO.NodeOutput:
+        validate_string(prompt, max_length=2000)
+        validate_images_aspect_ratio_closeness(first_frame, end_frame, min_rel=0.8, max_rel=1.25, strict=False)
+        payload = TaskCreationRequest(
+            model=model["model"],
+            prompt=prompt,
+            duration=model["duration"],
+            seed=seed,
+            resolution=model["resolution"],
+            audio=model["audio"],
+            images=[
+                (await upload_images_to_comfyapi(cls, frame, max_images=1, mime_type="image/png"))[0]
+                for frame in (first_frame, end_frame)
+            ],
+        )
+        results = await execute_task(cls, VIDU_START_END_VIDEO, payload)
+        return IO.NodeOutput(await download_url_to_video_output(results[0].url))
+
+
 class ViduExtension(ComfyExtension):
     @override
     async def get_node_list(self) -> list[type[IO.ComfyNode]]:
@@ -1497,6 +1706,7 @@ class ViduExtension(ComfyExtension):
             ViduMultiFrameVideoNode,
             Vidu3TextToVideoNode,
             Vidu3ImageToVideoNode,
+            Vidu3StartEndToVideoNode,
         ]
 
 

comfyui_version.py

@@ -1,3 +1,3 @@
 # This file is automatically generated by the build process when version is
 # updated in pyproject.toml.
-__version__ = "0.14.0"
+__version__ = "0.14.1"

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "ComfyUI"
-version = "0.14.0"
+version = "0.14.1"
 readme = "README.md"
 license = { file = "LICENSE" }
 requires-python = ">=3.10"

requirements.txt

@@ -1,5 +1,5 @@
 comfyui-frontend-package==1.39.14
-comfyui-workflow-templates==0.8.42
+comfyui-workflow-templates==0.8.43
 comfyui-embedded-docs==0.4.1
 torch
 torchsde