Merge branch 'comfyanonymous:master' into master

comfy/utils.py

@@ -78,8 +78,6 @@ def load_torch_file(ckpt, safe_load=False, device=None, return_metadata=False):
             pl_sd = torch.load(ckpt, map_location=device, weights_only=True, **torch_args)
         else:
             pl_sd = torch.load(ckpt, map_location=device, pickle_module=comfy.checkpoint_pickle)
-        if "global_step" in pl_sd:
-            logging.debug(f"Global Step: {pl_sd['global_step']}")
         if "state_dict" in pl_sd:
             sd = pl_sd["state_dict"]
         else:

comfy_api_nodes/nodes_kling.py

@@ -65,6 +65,12 @@ from comfy_api_nodes.apinode_utils import (
     download_url_to_image_tensor,
 )
 from comfy_api_nodes.mapper_utils import model_field_to_node_input
+from comfy_api_nodes.util.validation_utils import (
+    validate_image_dimensions,
+    validate_image_aspect_ratio,
+    validate_video_dimensions,
+    validate_video_duration,
+)
 from comfy_api.input.basic_types import AudioInput
 from comfy_api.input.video_types import VideoInput
 from comfy_api.input_impl import VideoFromFile
@@ -80,18 +86,16 @@ PATH_CHARACTER_IMAGE = f"/proxy/kling/{KLING_API_VERSION}/images/generations"
 PATH_VIRTUAL_TRY_ON = f"/proxy/kling/{KLING_API_VERSION}/images/kolors-virtual-try-on"
 PATH_IMAGE_GENERATIONS = f"/proxy/kling/{KLING_API_VERSION}/images/generations"
 
-
 MAX_PROMPT_LENGTH_T2V = 2500
 MAX_PROMPT_LENGTH_I2V = 500
 MAX_PROMPT_LENGTH_IMAGE_GEN = 500
 MAX_NEGATIVE_PROMPT_LENGTH_IMAGE_GEN = 200
 MAX_PROMPT_LENGTH_LIP_SYNC = 120
 
-# TODO: adjust based on tests
+AVERAGE_DURATION_T2V = 319
-AVERAGE_DURATION_T2V = 319  # 319,
+AVERAGE_DURATION_I2V = 164
-AVERAGE_DURATION_I2V = 164  # 164,
+AVERAGE_DURATION_LIP_SYNC = 455
-AVERAGE_DURATION_LIP_SYNC = 120
+AVERAGE_DURATION_VIRTUAL_TRY_ON = 19
-AVERAGE_DURATION_VIRTUAL_TRY_ON = 19  # 19,
 AVERAGE_DURATION_IMAGE_GEN = 32
 AVERAGE_DURATION_VIDEO_EFFECTS = 320
 AVERAGE_DURATION_VIDEO_EXTEND = 320
@@ -211,23 +215,8 @@ def validate_input_image(image: torch.Tensor) -> None:
 
     See: https://app.klingai.com/global/dev/document-api/apiReference/model/imageToVideo
     """
-    if len(image.shape) == 4:
+    validate_image_dimensions(image, min_width=300, min_height=300)
-        height, width = image.shape[1], image.shape[2]
+    validate_image_aspect_ratio(image, min_aspect_ratio=1 / 2.5, max_aspect_ratio=2.5)
-    elif len(image.shape) == 3:
-        height, width = image.shape[0], image.shape[1]
-    else:
-        raise ValueError("Invalid image tensor shape.")
-
-    # Ensure minimum resolution is met
-    if height < 300:
-        raise ValueError("Image height must be at least 300px")
-    if width < 300:
-        raise ValueError("Image width must be at least 300px")
-
-    # Ensure aspect ratio is within acceptable range
-    aspect_ratio = width / height
-    if aspect_ratio < 1 / 2.5 or aspect_ratio > 2.5:
-        raise ValueError("Image aspect ratio must be between 1:2.5 and 2.5:1")
 
 
 def get_camera_control_input_config(
@@ -1243,6 +1232,17 @@ class KlingLipSyncBase(KlingNodeBase):
     RETURN_TYPES = ("VIDEO", "STRING", "STRING")
     RETURN_NAMES = ("VIDEO", "video_id", "duration")
 
+    def validate_lip_sync_video(self, video: VideoInput):
+        """
+        Validates the input video adheres to the expectations of the Kling Lip Sync API:
+        - Video length does not exceed 10s and is not shorter than 2s
+        - Length and width dimensions should both be between 720px and 1920px
+
+        See: https://app.klingai.com/global/dev/document-api/apiReference/model/videoTolip
+        """
+        validate_video_dimensions(video, 720, 1920)
+        validate_video_duration(video, 2, 10)
+
     def validate_text(self, text: str):
         if not text:
             raise ValueError("Text is required")
@@ -1282,6 +1282,7 @@ class KlingLipSyncBase(KlingNodeBase):
     ) -> tuple[VideoFromFile, str, str]:
         if text:
             self.validate_text(text)
+        self.validate_lip_sync_video(video)
 
         # Upload video to Comfy API and get download URL
         video_url = upload_video_to_comfyapi(video, auth_kwargs=kwargs)
@@ -1352,7 +1353,7 @@ class KlingLipSyncAudioToVideoNode(KlingLipSyncBase):
             },
         }
 
-    DESCRIPTION = "Kling Lip Sync Audio to Video Node. Syncs mouth movements in a video file to the audio content of an audio file."
+    DESCRIPTION = "Kling Lip Sync Audio to Video Node. Syncs mouth movements in a video file to the audio content of an audio file. When using, ensure that the audio contains clearly distinguishable vocals and that the video contains a distinct face. The audio file should not be larger than 5MB. The video file should not be larger than 100MB, should have height/width between 720px and 1920px, and should be between 2s and 10s in length."
 
     def api_call(
         self,
@@ -1464,7 +1465,7 @@ class KlingLipSyncTextToVideoNode(KlingLipSyncBase):
             },
         }
 
-    DESCRIPTION = "Kling Lip Sync Text to Video Node. Syncs mouth movements in a video file to a text prompt."
+    DESCRIPTION = "Kling Lip Sync Text to Video Node. Syncs mouth movements in a video file to a text prompt. The video file should not be larger than 100MB, should have height/width between 720px and 1920px, and should be between 2s and 10s in length."
 
     def api_call(
         self,

comfy_api_nodes/util/validation_utils.py

@@ -0,0 +1,100 @@
+import logging
+from typing import Optional
+
+import torch
+from comfy_api.input.video_types import VideoInput
+
+
+def get_image_dimensions(image: torch.Tensor) -> tuple[int, int]:
+    if len(image.shape) == 4:
+        return image.shape[1], image.shape[2]
+    elif len(image.shape) == 3:
+        return image.shape[0], image.shape[1]
+    else:
+        raise ValueError("Invalid image tensor shape.")
+
+
+def validate_image_dimensions(
+    image: torch.Tensor,
+    min_width: Optional[int] = None,
+    max_width: Optional[int] = None,
+    min_height: Optional[int] = None,
+    max_height: Optional[int] = None,
+):
+    height, width = get_image_dimensions(image)
+
+    if min_width is not None and width < min_width:
+        raise ValueError(f"Image width must be at least {min_width}px, got {width}px")
+    if max_width is not None and width > max_width:
+        raise ValueError(f"Image width must be at most {max_width}px, got {width}px")
+    if min_height is not None and height < min_height:
+        raise ValueError(
+            f"Image height must be at least {min_height}px, got {height}px"
+        )
+    if max_height is not None and height > max_height:
+        raise ValueError(f"Image height must be at most {max_height}px, got {height}px")
+
+
+def validate_image_aspect_ratio(
+    image: torch.Tensor,
+    min_aspect_ratio: Optional[float] = None,
+    max_aspect_ratio: Optional[float] = None,
+):
+    width, height = get_image_dimensions(image)
+    aspect_ratio = width / height
+
+    if min_aspect_ratio is not None and aspect_ratio < min_aspect_ratio:
+        raise ValueError(
+            f"Image aspect ratio must be at least {min_aspect_ratio}, got {aspect_ratio}"
+        )
+    if max_aspect_ratio is not None and aspect_ratio > max_aspect_ratio:
+        raise ValueError(
+            f"Image aspect ratio must be at most {max_aspect_ratio}, got {aspect_ratio}"
+        )
+
+
+def validate_video_dimensions(
+    video: VideoInput,
+    min_width: Optional[int] = None,
+    max_width: Optional[int] = None,
+    min_height: Optional[int] = None,
+    max_height: Optional[int] = None,
+):
+    try:
+        width, height = video.get_dimensions()
+    except Exception as e:
+        logging.error("Error getting dimensions of video: %s", e)
+        return
+
+    if min_width is not None and width < min_width:
+        raise ValueError(f"Video width must be at least {min_width}px, got {width}px")
+    if max_width is not None and width > max_width:
+        raise ValueError(f"Video width must be at most {max_width}px, got {width}px")
+    if min_height is not None and height < min_height:
+        raise ValueError(
+            f"Video height must be at least {min_height}px, got {height}px"
+        )
+    if max_height is not None and height > max_height:
+        raise ValueError(f"Video height must be at most {max_height}px, got {height}px")
+
+
+def validate_video_duration(
+    video: VideoInput,
+    min_duration: Optional[float] = None,
+    max_duration: Optional[float] = None,
+):
+    try:
+        duration = video.get_duration()
+    except Exception as e:
+        logging.error("Error getting duration of video: %s", e)
+        return
+
+    epsilon = 0.0001
+    if min_duration is not None and min_duration - epsilon > duration:
+        raise ValueError(
+            f"Video duration must be at least {min_duration}s, got {duration}s"
+        )
+    if max_duration is not None and duration > max_duration + epsilon:
+        raise ValueError(
+            f"Video duration must be at most {max_duration}s, got {duration}s"
+        )

comfy_extras/nodes_cond.py

@@ -31,6 +31,7 @@ class T5TokenizerOptions:
             }
         }
 
+    CATEGORY = "_for_testing/conditioning"
     RETURN_TYPES = ("CLIP",)
     FUNCTION = "set_options"
 

comfy_extras/nodes_images.py

@@ -13,6 +13,7 @@ import os
 import re
 from io import BytesIO
 from inspect import cleandoc
+import torch
 
 from comfy.comfy_types import FileLocator
 
@@ -74,6 +75,24 @@ class ImageFromBatch:
         s = s_in[batch_index:batch_index + length].clone()
         return (s,)
 
+
+class ImageAddNoise:
+    @classmethod
+    def INPUT_TYPES(s):
+        return {"required": { "image": ("IMAGE",),
+                              "seed": ("INT", {"default": 0, "min": 0, "max": 0xffffffffffffffff, "control_after_generate": True, "tooltip": "The random seed used for creating the noise."}),
+                              "strength": ("FLOAT", {"default": 0.5, "min": 0.0, "max": 1.0, "step": 0.01}),
+                              }}
+    RETURN_TYPES = ("IMAGE",)
+    FUNCTION = "repeat"
+
+    CATEGORY = "image"
+
+    def repeat(self, image, seed, strength):
+        generator = torch.manual_seed(seed)
+        s = torch.clip((image + strength * torch.randn(image.size(), generator=generator, device="cpu").to(image)), min=0.0, max=1.0)
+        return (s,)
+
 class SaveAnimatedWEBP:
     def __init__(self):
         self.output_dir = folder_paths.get_output_directory()
@@ -295,6 +314,7 @@ NODE_CLASS_MAPPINGS = {
     "ImageCrop": ImageCrop,
     "RepeatImageBatch": RepeatImageBatch,
     "ImageFromBatch": ImageFromBatch,
+    "ImageAddNoise": ImageAddNoise,
     "SaveAnimatedWEBP": SaveAnimatedWEBP,
     "SaveAnimatedPNG": SaveAnimatedPNG,
     "SaveSVGNode": SaveSVGNode,

comfy_extras/nodes_string.py

@@ -8,7 +8,8 @@ class StringConcatenate():
         return {
             "required": {
                 "string_a": (IO.STRING, {"multiline": True}),
-                "string_b": (IO.STRING, {"multiline": True})
+                "string_b": (IO.STRING, {"multiline": True}),
+                "delimiter": (IO.STRING, {"multiline": False, "default": ""})
             }
         }
 
@@ -16,8 +17,8 @@ class StringConcatenate():
     FUNCTION = "execute"
     CATEGORY = "utils/string"
 
-    def execute(self, string_a, string_b, **kwargs):
+    def execute(self, string_a, string_b, delimiter, **kwargs):
-        return string_a + string_b,
+        return delimiter.join((string_a, string_b)),
 
 class StringSubstring():
     @classmethod

nodes.py

@@ -1940,7 +1940,7 @@ class ImagePadForOutpaint:
 
         mask[top:top + d2, left:left + d3] = t
 
-        return (new_image, mask)
+        return (new_image, mask.unsqueeze(0))
 
 
 NODE_CLASS_MAPPINGS = {