Fix reshaping lora application (#14031)

* ModelPatcherDyanmic: purge stale vbar allocs on force cast

* ModelPatcherDynamic: restore backups before load

If doing a clean reload, mutative changes (lora application) could be
applied on-top of the already loaded weight. Restore from backup
unconditionally so that the new load is clean.

comfy/model_patcher.py

@@ -1613,6 +1613,16 @@ class ModelPatcherDynamic(ModelPatcher):
         #use all ModelPatcherDynamic this is ignored and its all done dynamically.
         return super().memory_required(input_shape=input_shape) * 1.3 + (1024 ** 3)
 
+    def restore_loaded_backups(self):
+        restored = self.model.model_loaded_weight_memory
+        for key in list(self.backup.keys()):
+            bk = self.backup.pop(key)
+            comfy.utils.set_attr_param(self.model, key, bk.weight)
+        for key in list(self.backup_buffers.keys()):
+            comfy.utils.set_attr_buffer(self.model, key, self.backup_buffers.pop(key))
+        self.model.model_loaded_weight_memory = 0
+        return restored
+
 
     def load(self, device_to=None, lowvram_model_memory=0, force_patch_weights=False, full_load=False, dirty=False):
 
@@ -1629,7 +1639,7 @@ class ModelPatcherDynamic(ModelPatcher):
 
         num_patches = 0
         allocated_size = 0
-        self.model.model_loaded_weight_memory = 0
+        self.restore_loaded_backups()
 
         with self.use_ejected():
             self.unpatch_hooks()
@@ -1716,6 +1726,9 @@ class ModelPatcherDynamic(ModelPatcher):
                         force_load=True
 
                     if force_load:
+                        if hasattr(m, "_v"):
+                            comfy_aimdo.model_vbar.vbar_unpin(m._v)
+                            delattr(m, "_v")
                         force_load_param(self, "weight", device_to)
                         force_load_param(self, "bias", device_to)
                     else:
@@ -1773,13 +1786,7 @@ class ModelPatcherDynamic(ModelPatcher):
         freed = 0 if vbar is None else vbar.free_memory(memory_to_free)
 
         if freed < memory_to_free:
-            for key in list(self.backup.keys()):
-                bk = self.backup.pop(key)
-                comfy.utils.set_attr_param(self.model, key, bk.weight)
-            for key in list(self.backup_buffers.keys()):
-                comfy.utils.set_attr_buffer(self.model, key, self.backup_buffers.pop(key))
-            freed += self.model.model_loaded_weight_memory
-            self.model.model_loaded_weight_memory = 0
+            freed += self.restore_loaded_backups()
 
         return freed
 

comfy/utils.py

@@ -1019,10 +1019,11 @@ def bislerp(samples, width, height):
 
 def lanczos(samples, width, height):
     #the below API is strict and expects grayscale to be squeezed
-    samples = samples.squeeze(1) if samples.shape[1] == 1 else samples.movedim(1, -1)
+    if samples.ndim == 4:
+        samples = samples.squeeze(1) if samples.shape[1] == 1 else samples.movedim(1, -1)
     images = [Image.fromarray(np.clip(255. * image.cpu().numpy(), 0, 255).astype(np.uint8)) for image in samples]
     images = [image.resize((width, height), resample=Image.Resampling.LANCZOS) for image in images]
-    images = [torch.from_numpy(np.array(image).astype(np.float32) / 255.0).movedim(-1, 0) for image in images]
+    images = [torch.from_numpy(t).movedim(-1, 0) if (t := np.array(image).astype(np.float32) / 255.0).ndim == 3 else torch.from_numpy(t) for image in images]
     result = torch.stack(images)
     return result.to(samples.device, samples.dtype)
 

comfy_extras/nodes_audio.py

@@ -543,7 +543,7 @@ class AudioConcat(IO.ComfyNode):
         return IO.Schema(
             node_id="AudioConcat",
             search_aliases=["join audio", "combine audio", "append audio"],
-            display_name="Audio Concat",
+            display_name="Concatenate Audio",
             description="Concatenates the audio1 to audio2 in the specified direction.",
             category="audio",
             inputs=[
@@ -597,7 +597,7 @@ class AudioMerge(IO.ComfyNode):
         return IO.Schema(
             node_id="AudioMerge",
             search_aliases=["mix audio", "overlay audio", "layer audio"],
-            display_name="Audio Merge",
+            display_name="Merge Audio",
             description="Combine two audio tracks by overlaying their waveforms.",
             category="audio",
             inputs=[
@@ -667,8 +667,9 @@ class AudioAdjustVolume(IO.ComfyNode):
         return IO.Schema(
             node_id="AudioAdjustVolume",
             search_aliases=["audio gain", "loudness", "audio level"],
-            display_name="Audio Adjust Volume",
+            display_name="Adjust Audio Volume",
             category="audio",
+            description="Adjust the volume of the audio by a specified amount in decibels (dB).",
             inputs=[
                 IO.Audio.Input("audio"),
                 IO.Int.Input(

comfy_extras/nodes_dataset.py

@@ -47,8 +47,10 @@ class LoadImageDataSetFromFolderNode(io.ComfyNode):
     def define_schema(cls):
         return io.Schema(
             node_id="LoadImageDataSetFromFolder",
-            display_name="Load Image Dataset from Folder",
-            category="dataset",
+            search_aliases=["load folder", "load from folder", "load dataset", "load images", "import dataset"],
+            display_name="Load Image (from Folder)",
+            category="image",
+            description="Load a dataset of images from a specified folder and return a list of images. Supported formats: PNG, JPG, JPEG, WEBP.",
             is_experimental=True,
             inputs=[
                 io.Combo.Input(
@@ -84,14 +86,16 @@ class LoadImageTextDataSetFromFolderNode(io.ComfyNode):
     def define_schema(cls):
         return io.Schema(
             node_id="LoadImageTextDataSetFromFolder",
-            display_name="Load Image and Text Dataset from Folder",
-            category="dataset",
+            search_aliases=["load folder", "load from folder", "load dataset", "load images", "import dataset"],
+            display_name="Load Image-Text (from Folder)",
+            category="image",
+            description="Load a dataset of pairs of images and text captions from a specified folder and return them as a list. Supported formats: PNG, JPG, JPEG, WEBP.",
             is_experimental=True,
             inputs=[
                 io.Combo.Input(
                     "folder",
                     options=folder_paths.get_input_subfolders(),
-                    tooltip="The folder to load images from.",
+                    tooltip="The folder to load images and text captions from.",
                 )
             ],
             outputs=[
@@ -206,8 +210,10 @@ class SaveImageDataSetToFolderNode(io.ComfyNode):
     def define_schema(cls):
         return io.Schema(
             node_id="SaveImageDataSetToFolder",
-            display_name="Save Image Dataset to Folder",
-            category="dataset",
+            search_aliases=["save folder", "save to folder", "save dataset", "save images", "export dataset"],
+            display_name="Save Image (to Folder) (DEPRECATED)",
+            category="image",
+            description="Save a dataset of images to a specified folder. Supported formats: PNG.",
             is_experimental=True,
             is_output_node=True,
             is_input_list=True,  # Receive images as list
@@ -226,6 +232,7 @@ class SaveImageDataSetToFolderNode(io.ComfyNode):
                 ),
             ],
             outputs=[],
+            is_deprecated=True,  # This node is redundant and superseded by existing Save Image nodes where the target folder can be specified in the filename_prefix
         )
 
     @classmethod
@@ -246,14 +253,20 @@ class SaveImageTextDataSetToFolderNode(io.ComfyNode):
     def define_schema(cls):
         return io.Schema(
             node_id="SaveImageTextDataSetToFolder",
-            display_name="Save Image and Text Dataset to Folder",
-            category="dataset",
+            search_aliases=["save folder", "save to folder", "save dataset", "save images", "save text", "export dataset"],
+            display_name="Save Image-Text (to Folder)",
+            category="image",
+            description="Save a dataset of pairs of images and text captions to a specified folder. Images are saved as PNG files and captions are saved as TXT files with the same filename_prefix.",
             is_experimental=True,
             is_output_node=True,
             is_input_list=True,  # Receive both images and texts as lists
             inputs=[
                 io.Image.Input("images", tooltip="List of images to save."),
-                io.String.Input("texts", tooltip="List of text captions to save."),
+                io.String.Input("texts",
+                    optional=True,
+                    force_input=True,
+                    tooltip="List of text captions to save."
+                ),
                 io.String.Input(
                     "folder_name",
                     default="dataset",
@@ -270,7 +283,7 @@ class SaveImageTextDataSetToFolderNode(io.ComfyNode):
         )
 
     @classmethod
-    def execute(cls, images, texts, folder_name, filename_prefix):
+    def execute(cls, images, folder_name, filename_prefix, texts=None):
         # Extract scalar values
         folder_name = folder_name[0]
         filename_prefix = filename_prefix[0]
@@ -279,11 +292,12 @@ class SaveImageTextDataSetToFolderNode(io.ComfyNode):
         saved_files = save_images_to_folder(images, output_dir, filename_prefix)
 
         # Save captions
-        for idx, (filename, caption) in enumerate(zip(saved_files, texts)):
-            caption_filename = filename.replace(".png", ".txt")
-            caption_path = os.path.join(output_dir, caption_filename)
-            with open(caption_path, "w", encoding="utf-8") as f:
-                f.write(caption)
+        if texts:
+            for idx, (filename, caption) in enumerate(zip(saved_files, texts)):
+                caption_filename = filename.replace(".png", ".txt")
+                caption_path = os.path.join(output_dir, caption_filename)
+                with open(caption_path, "w", encoding="utf-8") as f:
+                    f.write(caption)
 
         logging.info(f"Saved {len(saved_files)} images and captions to {output_dir}.")
         return io.NodeOutput()
@@ -314,11 +328,13 @@ class ImageProcessingNode(io.ComfyNode):
 
     Child classes should set:
         node_id: Unique node identifier (required)
+        search_aliases: List of search aliases (optional)
         display_name: Display name (optional, defaults to node_id)
         description: Node description (optional)
         extra_inputs: List of additional io.Input objects beyond "images" (optional)
         is_group_process: None (auto-detect), True (group), or False (individual) (optional)
         is_output_list: True (list output) or False (single output) (optional, default True)
+        is_deprecated: True if the node is deprecated (optional, default False)
 
     Child classes must implement ONE of:
         _process(cls, image, **kwargs) -> tensor  (for single-item processing)
@@ -326,12 +342,13 @@ class ImageProcessingNode(io.ComfyNode):
     """
 
     node_id = None
+    search_aliases = []
     display_name = None
     description = None
     extra_inputs = []
     is_group_process = None  # None = auto-detect, True/False = explicit
     is_output_list = None  # None = auto-detect based on processing mode
-
+    is_deprecated = False
     @classmethod
     def _detect_processing_mode(cls):
         """Detect whether this node uses group or individual processing.
@@ -402,8 +419,10 @@ class ImageProcessingNode(io.ComfyNode):
 
         return io.Schema(
             node_id=cls.node_id,
+            search_aliases=cls.search_aliases,
             display_name=cls.display_name or cls.node_id,
-            category="dataset/image",
+            category=cls.category,
+            description=cls.description,
             is_experimental=True,
             is_input_list=is_group,  # True for group, False for individual
             inputs=inputs,
@@ -472,11 +491,13 @@ class TextProcessingNode(io.ComfyNode):
 
     Child classes should set:
         node_id: Unique node identifier (required)
+        search_aliases: List of search aliases (optional)
         display_name: Display name (optional, defaults to node_id)
         description: Node description (optional)
         extra_inputs: List of additional io.Input objects beyond "texts" (optional)
         is_group_process: None (auto-detect), True (group), or False (individual) (optional)
         is_output_list: True (list output) or False (single output) (optional, default True)
+        is_deprecated: True if the node is deprecated (optional, default False)
 
     Child classes must implement ONE of:
         _process(cls, text, **kwargs) -> str  (for single-item processing)
@@ -484,12 +505,13 @@ class TextProcessingNode(io.ComfyNode):
     """
 
     node_id = None
+    search_aliases = []
     display_name = None
     description = None
     extra_inputs = []
     is_group_process = None  # None = auto-detect, True/False = explicit
     is_output_list = None  # None = auto-detect based on processing mode
-
+    is_deprecated = False
     @classmethod
     def _detect_processing_mode(cls):
         """Detect whether this node uses group or individual processing.
@@ -627,15 +649,17 @@ class TextProcessingNode(io.ComfyNode):
 
 class ResizeImagesByShorterEdgeNode(ImageProcessingNode):
     node_id = "ResizeImagesByShorterEdge"
-    display_name = "Resize Images by Shorter Edge"
-    description = "Resize images so that the shorter edge matches the specified length while preserving aspect ratio."
+    display_name = "Resize Images by Shorter Edge (DEPRECATED)"
+    category = "image/transform"
+    description = "Resize images so that the shorter edge matches the specified dimension while preserving aspect ratio."
+    is_deprecated = True  # This node is superseded by Resize Image/Mask with resize_type = scale shorter dimension
     extra_inputs = [
         io.Int.Input(
             "shorter_edge",
             default=512,
             min=1,
             max=8192,
-            tooltip="Target length for the shorter edge.",
+            tooltip="Target dimension for the shorter edge.",
         ),
     ]
 
@@ -655,15 +679,17 @@ class ResizeImagesByShorterEdgeNode(ImageProcessingNode):
 
 class ResizeImagesByLongerEdgeNode(ImageProcessingNode):
     node_id = "ResizeImagesByLongerEdge"
-    display_name = "Resize Images by Longer Edge"
-    description = "Resize images so that the longer edge matches the specified length while preserving aspect ratio."
+    display_name = "Resize Images by Longer Edge (DEPRECATED)"
+    category = "image/transform"
+    description = "Resize images so that the longer edge matches the specified dimension while preserving aspect ratio."
+    is_deprecated = True  # This node is superseded by Resize Image/Mask with resize_type = scale longer dimension
     extra_inputs = [
         io.Int.Input(
             "longer_edge",
             default=1024,
             min=1,
             max=8192,
-            tooltip="Target length for the longer edge.",
+            tooltip="Target dimension for the longer edge.",
         ),
     ]
 
@@ -686,8 +712,10 @@ class ResizeImagesByLongerEdgeNode(ImageProcessingNode):
 
 class CenterCropImagesNode(ImageProcessingNode):
     node_id = "CenterCropImages"
-    display_name = "Center Crop Images"
-    description = "Center crop all images to the specified dimensions."
+    search_aliases=["crop", "cut", "trim"]
+    display_name="Crop Image (Center)"
+    category="image/transform"
+    description = "Center crop an image to the specified dimensions."
     extra_inputs = [
         io.Int.Input("width", default=512, min=1, max=8192, tooltip="Crop width."),
         io.Int.Input("height", default=512, min=1, max=8192, tooltip="Crop height."),
@@ -706,10 +734,11 @@ class CenterCropImagesNode(ImageProcessingNode):
 
 class RandomCropImagesNode(ImageProcessingNode):
     node_id = "RandomCropImages"
-    display_name = "Random Crop Images"
-    description = (
-        "Randomly crop all images to the specified dimensions (for data augmentation)."
-    )
+    search_aliases=["crop", "cut", "trim"]
+    display_name = "Crop Image (Random)"
+    category="image/transform"
+    description = "Randomly crop an image to the specified dimensions."
+
     extra_inputs = [
         io.Int.Input("width", default=512, min=1, max=8192, tooltip="Crop width."),
         io.Int.Input("height", default=512, min=1, max=8192, tooltip="Crop height."),
@@ -734,7 +763,9 @@ class RandomCropImagesNode(ImageProcessingNode):
 
 class NormalizeImagesNode(ImageProcessingNode):
     node_id = "NormalizeImages"
-    display_name = "Normalize Images"
+    search_aliases=["normalize", "normalize colors"]
+    display_name = "Normalize Image Colors"
+    category = "image/color"
     description = "Normalize images using mean and standard deviation."
     extra_inputs = [
         io.Float.Input(
@@ -762,8 +793,10 @@ class NormalizeImagesNode(ImageProcessingNode):
 
 class AdjustBrightnessNode(ImageProcessingNode):
     node_id = "AdjustBrightness"
+    search_aliases=["brightness"]
     display_name = "Adjust Brightness"
-    description = "Adjust brightness of all images."
+    category="image/adjustments"
+    description = "Adjust the brightness of an image."
     extra_inputs = [
         io.Float.Input(
             "factor",
@@ -781,8 +814,10 @@ class AdjustBrightnessNode(ImageProcessingNode):
 
 class AdjustContrastNode(ImageProcessingNode):
     node_id = "AdjustContrast"
+    search_aliases=["contrast"]
     display_name = "Adjust Contrast"
-    description = "Adjust contrast of all images."
+    category="image/adjustments"
+    description = "Adjust the contrast of an image."
     extra_inputs = [
         io.Float.Input(
             "factor",
@@ -800,8 +835,10 @@ class AdjustContrastNode(ImageProcessingNode):
 
 class ShuffleDatasetNode(ImageProcessingNode):
     node_id = "ShuffleDataset"
-    display_name = "Shuffle Image Dataset"
-    description = "Randomly shuffle the order of images in the dataset."
+    search_aliases=["shuffle", "randomize", "mix"]
+    display_name = "Shuffle Images List"
+    category = "image/batch"
+    description = "Randomly shuffle the order of images in a list."
     is_group_process = True  # Requires full list to shuffle
     extra_inputs = [
         io.Int.Input(
@@ -823,13 +860,15 @@ class ShuffleImageTextDatasetNode(io.ComfyNode):
     def define_schema(cls):
         return io.Schema(
             node_id="ShuffleImageTextDataset",
-            display_name="Shuffle Image-Text Dataset",
-            category="dataset/image",
+            search_aliases=["shuffle", "randomize", "mix"],
+            display_name = "Shuffle Pairs of Image-Text",
+            category = "image/batch",
+            description = "Randomly shuffle the order of pairs of image-text in a list.",
             is_experimental=True,
             is_input_list=True,
             inputs=[
                 io.Image.Input("images", tooltip="List of images to shuffle."),
-                io.String.Input("texts", tooltip="List of texts to shuffle."),
+                io.String.Input("texts", tooltip="List of texts to shuffle.", force_input=True),
                 io.Int.Input(
                     "seed",
                     default=0,
@@ -865,8 +904,11 @@ class ShuffleImageTextDatasetNode(io.ComfyNode):
 
 class TextToLowercaseNode(TextProcessingNode):
     node_id = "TextToLowercase"
-    display_name = "Text to Lowercase"
-    description = "Convert all texts to lowercase."
+    search_aliases=["lowercase"]
+    display_name = "Convert Text to Lowercase (DEPRECATED)"
+    category = "text"
+    description = "Convert text to lowercase."
+    is_deprecated = True  # This node is superseded by the Convert Text Case node
 
     @classmethod
     def _process(cls, text):
@@ -875,8 +917,11 @@ class TextToLowercaseNode(TextProcessingNode):
 
 class TextToUppercaseNode(TextProcessingNode):
     node_id = "TextToUppercase"
-    display_name = "Text to Uppercase"
-    description = "Convert all texts to uppercase."
+    search_aliases=["uppercase"]
+    display_name = "Convert Text to Uppercase (DEPRECATED)"
+    category = "text"
+    description = "Convert text to uppercase."
+    is_deprecated = True  # This node is superseded by the Convert Text Case node
 
     @classmethod
     def _process(cls, text):
@@ -885,8 +930,10 @@ class TextToUppercaseNode(TextProcessingNode):
 
 class TruncateTextNode(TextProcessingNode):
     node_id = "TruncateText"
+    search_aliases=["truncate", "cut", "shorten"]
     display_name = "Truncate Text"
-    description = "Truncate all texts to a maximum length."
+    category = "text"
+    description = "Truncate text to a maximum length."
     extra_inputs = [
         io.Int.Input(
             "max_length", default=77, min=1, max=10000, tooltip="Maximum text length."
@@ -900,8 +947,10 @@ class TruncateTextNode(TextProcessingNode):
 
 class AddTextPrefixNode(TextProcessingNode):
     node_id = "AddTextPrefix"
-    display_name = "Add Text Prefix"
+    display_name = "Add Text Prefix (DEPRECATED)"
+    category = "text"
     description = "Add a prefix to all texts."
+    is_deprecated = True  # This node is superseded by the Concatenate Text node
     extra_inputs = [
         io.String.Input("prefix", default="", tooltip="Prefix to add."),
     ]
@@ -913,8 +962,10 @@ class AddTextPrefixNode(TextProcessingNode):
 
 class AddTextSuffixNode(TextProcessingNode):
     node_id = "AddTextSuffix"
-    display_name = "Add Text Suffix"
+    display_name = "Add Text Suffix (DEPRECATED)"
+    category = "text"
     description = "Add a suffix to all texts."
+    is_deprecated = True  # This node is superseded by the Concatenate Text node
     extra_inputs = [
         io.String.Input("suffix", default="", tooltip="Suffix to add."),
     ]
@@ -926,8 +977,10 @@ class AddTextSuffixNode(TextProcessingNode):
 
 class ReplaceTextNode(TextProcessingNode):
     node_id = "ReplaceText"
-    display_name = "Replace Text"
+    display_name = "Replace Text (DEPRECATED)"
+    category = "text"
     description = "Replace text in all texts."
+    is_deprecated = True  # This node is superseded by the other Replace Text node
     extra_inputs = [
         io.String.Input("find", default="", tooltip="Text to find."),
         io.String.Input("replace", default="", tooltip="Text to replace with."),
@@ -940,8 +993,10 @@ class ReplaceTextNode(TextProcessingNode):
 
 class StripWhitespaceNode(TextProcessingNode):
     node_id = "StripWhitespace"
-    display_name = "Strip Whitespace"
+    display_name = "Strip Whitespace (DEPRECATED)"
+    category = "text"
     description = "Strip leading and trailing whitespace from all texts."
+    is_deprecated = True  # This node is superseded by the Trim Text node
 
     @classmethod
     def _process(cls, text):
@@ -952,11 +1007,13 @@ class StripWhitespaceNode(TextProcessingNode):
 
 
 class ImageDeduplicationNode(ImageProcessingNode):
-    """Remove duplicate or very similar images from the dataset using perceptual hashing."""
+    """Remove duplicate or very similar images from a list using perceptual hashing."""
 
     node_id = "ImageDeduplication"
-    display_name = "Image Deduplication"
-    description = "Remove duplicate or very similar images from the dataset."
+    search_aliases=["deduplicate", "remove duplicates", "similarity filter"]
+    display_name = "Deduplicate Images"
+    category = "image/batch"
+    description = "Remove duplicate or very similar images from a list."
     is_group_process = True  # Requires full list to compare images
     extra_inputs = [
         io.Float.Input(
@@ -1026,7 +1083,9 @@ class ImageGridNode(ImageProcessingNode):
     """Combine multiple images into a single grid/collage."""
 
     node_id = "ImageGrid"
-    display_name = "Image Grid"
+    search_aliases=["grid", "collage", "combine"]
+    display_name = "Make Image Grid"
+    category="image/batch"
     description = "Arrange multiple images into a grid layout."
     is_group_process = True  # Requires full list to create grid
     is_output_list = False  # Outputs single grid image
@@ -1102,9 +1161,12 @@ class MergeImageListsNode(ImageProcessingNode):
     """Merge multiple image lists into a single list."""
 
     node_id = "MergeImageLists"
-    display_name = "Merge Image Lists"
+    search_aliases=["list", "merge list", "make list"]
+    display_name = "Merge Image Lists (DEPRECATED)"
+    category = "image/batch"
     description = "Concatenate multiple image lists into one."
     is_group_process = True  # Receives images as list
+    is_deprecated = True  # This node is superseded by the Create List node
 
     @classmethod
     def _group_process(cls, images):
@@ -1119,9 +1181,11 @@ class MergeTextListsNode(TextProcessingNode):
     """Merge multiple text lists into a single list."""
 
     node_id = "MergeTextLists"
-    display_name = "Merge Text Lists"
+    display_name = "Merge Text Lists (DEPRECATED)"
+    category = "text"
     description = "Concatenate multiple text lists into one."
     is_group_process = True  # Receives texts as list
+    is_deprecated = True  # This node is superseded by the Create List node
 
     @classmethod
     def _group_process(cls, texts):
@@ -1142,8 +1206,10 @@ class ResolutionBucket(io.ComfyNode):
     def define_schema(cls):
         return io.Schema(
             node_id="ResolutionBucket",
+            search_aliases=["bucket by resolution", "group by resolution", "batch by resolution"],
             display_name="Resolution Bucket",
-            category="dataset",
+            category="training",
+            description="Group latents and conditionings into buckets",
             is_experimental=True,
             is_input_list=True,
             inputs=[
@@ -1236,7 +1302,8 @@ class MakeTrainingDataset(io.ComfyNode):
             node_id="MakeTrainingDataset",
             search_aliases=["encode dataset"],
             display_name="Make Training Dataset",
-            category="dataset",
+            category="training",
+            description="Encode images with VAE and texts with CLIP to create a training dataset of latents and conditionings.",
             is_experimental=True,
             is_input_list=True,  # images and texts as lists
             inputs=[
@@ -1251,6 +1318,7 @@ class MakeTrainingDataset(io.ComfyNode):
                     "texts",
                     optional=True,
                     tooltip="List of text captions. Can be length n (matching images), 1 (repeated for all), or omitted (uses empty string).",
+                    force_input=True
                 ),
             ],
             outputs=[
@@ -1320,9 +1388,10 @@ class SaveTrainingDataset(io.ComfyNode):
     def define_schema(cls):
         return io.Schema(
             node_id="SaveTrainingDataset",
-            search_aliases=["export training data"],
+            search_aliases=["export dataset", "save dataset"],
             display_name="Save Training Dataset",
-            category="dataset",
+            category="training",
+            description="Save encoded training dataset (latents + conditioning) to disk for efficient loading during training.",
             is_experimental=True,
             is_output_node=True,
             is_input_list=True,  # Receive lists
@@ -1424,7 +1493,8 @@ class LoadTrainingDataset(io.ComfyNode):
             node_id="LoadTrainingDataset",
             search_aliases=["import dataset", "training data"],
             display_name="Load Training Dataset",
-            category="dataset",
+            category="training",
+            description="Load encoded training dataset (latents + conditioning) from disk for use in training.",
             is_experimental=True,
             inputs=[
                 io.String.Input(

comfy_extras/nodes_hunyuan3d.py

@@ -419,15 +419,17 @@ class VoxelToMeshBasic(IO.ComfyNode):
     def define_schema(cls):
         return IO.Schema(
             node_id="VoxelToMeshBasic",
-            display_name="Voxel to Mesh (Basic)",
+            display_name="Voxel to Mesh (Basic) (DEPRECATED)",
             category="3d",
+            description="Converts a voxel grid to a mesh.",
+            is_deprecated=True, # This node is superseded by the Voxel To Mesh node
             inputs=[
                 IO.Voxel.Input("voxel"),
                 IO.Float.Input("threshold", default=0.6, min=-1.0, max=1.0, step=0.01),
             ],
             outputs=[
                 IO.Mesh.Output(),
-            ]
+            ],
         )
 
     @classmethod
@@ -453,9 +455,10 @@ class VoxelToMesh(IO.ComfyNode):
             node_id="VoxelToMesh",
             display_name="Voxel to Mesh",
             category="3d",
+            description="Converts a voxel grid to a mesh.",
             inputs=[
                 IO.Voxel.Input("voxel"),
-                IO.Combo.Input("algorithm", options=["surface net", "basic"], advanced=True),
+                IO.Combo.Input("algorithm", options=["surface net", "basic"]),
                 IO.Float.Input("threshold", default=0.6, min=-1.0, max=1.0, step=0.01),
             ],
             outputs=[

comfy_extras/nodes_images.py

@@ -55,9 +55,10 @@ class ImageCropV2(IO.ComfyNode):
     def define_schema(cls):
         return IO.Schema(
             node_id="ImageCropV2",
-            search_aliases=["trim"],
+            search_aliases=["crop", "cut", "trim"],
             display_name="Crop Image",
             category="image/transform",
+            description = "Crop an image to the specified dimensions.",
             essentials_category="Image Tools",
             has_intermediate_output=True,
             inputs=[

comfy_extras/nodes_lt_audio.py

@@ -11,8 +11,8 @@ class LTXVAudioVAELoader(io.ComfyNode):
     def define_schema(cls) -> io.Schema:
         return io.Schema(
             node_id="LTXVAudioVAELoader",
-            display_name="LTXV Audio VAE Loader",
-            category="audio",
+            display_name="Load LTXV Audio VAE",
+            category="loaders",
             inputs=[
                 io.Combo.Input(
                     "ckpt_name",
@@ -40,7 +40,7 @@ class LTXVAudioVAEEncode(VAEEncodeAudio):
         return io.Schema(
             node_id="LTXVAudioVAEEncode",
             display_name="LTXV Audio VAE Encode",
-            category="audio",
+            category="latent/audio",
             inputs=[
                 io.Audio.Input("audio", tooltip="The audio to be encoded."),
                 io.Vae.Input(
@@ -63,7 +63,7 @@ class LTXVAudioVAEDecode(io.ComfyNode):
         return io.Schema(
             node_id="LTXVAudioVAEDecode",
             display_name="LTXV Audio VAE Decode",
-            category="audio",
+            category="latent/audio",
             inputs=[
                 io.Latent.Input("samples", tooltip="The latent to be decoded."),
                 io.Vae.Input(

comfy_extras/nodes_mediapipe.py

@@ -28,7 +28,7 @@ from comfy_extras.mediapipe.face_landmarker import FaceLandmarker
 from comfy_extras.mediapipe.face_geometry import transformation_matrix_from_detection
 
 
-FaceLandmarkerType = io.Custom("FACE_LANDMARKER")
+FaceDetectionType = io.Custom("FACE_DETECTION_MODEL")
 FaceLandmarksType = io.Custom("FACE_LANDMARKS")
 
 _CANONICAL_KEYS = ("canonical_vertices", "procrustes_indices", "procrustes_weights")
@@ -204,18 +204,19 @@ class LoadMediaPipeFaceLandmarker(io.ComfyNode):
     def define_schema(cls):
         return io.Schema(
             node_id="LoadMediaPipeFaceLandmarker",
-            display_name="Load MediaPipe Face Landmarker",
+            search_aliases=["face", "facial", "mediapipe", "face landmark", "face mesh", "blazeface", "face detection"],
+            display_name="Load Face Detection Model (MediaPipe)",
             category="loaders",
             inputs=[
-                io.Combo.Input("model_name", options=folder_paths.get_filename_list("mediapipe"),
-                               tooltip="Face Landmarker safetensors from models/mediapipe/."),
+                io.Combo.Input("model_name", options=folder_paths.get_filename_list("detection"),
+                               tooltip="Face detection model from models/detection/."),
             ],
-            outputs=[FaceLandmarkerType.Output()],
+            outputs=[FaceDetectionType.Output()],
         )
 
     @classmethod
     def execute(cls, model_name) -> io.NodeOutput:
-        sd = comfy.utils.load_torch_file(folder_paths.get_full_path_or_raise("mediapipe", model_name), safe_load=True)
+        sd = comfy.utils.load_torch_file(folder_paths.get_full_path_or_raise("detection", model_name), safe_load=True)
         wrapper = FaceLandmarkerModel(sd)
         return io.NodeOutput(wrapper)
 
@@ -234,10 +235,12 @@ class MediaPipeFaceLandmarker(io.ComfyNode):
     def define_schema(cls):
         return io.Schema(
             node_id="MediaPipeFaceLandmarker",
-            display_name="MediaPipe Face Landmarker",
+            search_aliases=["face", "facial", "mediapipe", "face landmark", "face mesh", "blazeface", "face detection"],
+            display_name="Detect Face Landmarks (MediaPipe)",
             category="image/detection",
+            description="Detects facial landmarks using MediaPipe model.",
             inputs=[
-                FaceLandmarkerType.Input("face_landmarker"),
+                FaceDetectionType.Input("face_detection_model"),
                 io.Image.Input("image"),
                 io.Combo.Input("detector_variant", options=["short", "full", "both"], default="short",
                                tooltip="Face detector range. 'short' is tuned for close-up faces "
@@ -261,9 +264,9 @@ class MediaPipeFaceLandmarker(io.ComfyNode):
         )
 
     @classmethod
-    def execute(cls, face_landmarker, image, detector_variant, num_faces, min_confidence,
+    def execute(cls, face_detection_model, image, detector_variant, num_faces, min_confidence,
                 missing_frame_fallback) -> io.NodeOutput:
-        canonical = face_landmarker.canonical_data
+        canonical = face_detection_model.canonical_data
         img_np = _image_to_uint8(image)
         B, H, W = img_np.shape[:3]
         chunk = 16
@@ -276,7 +279,7 @@ class MediaPipeFaceLandmarker(io.ComfyNode):
             with tqdm(total=B, desc=f"MediaPipe Face Landmarker ({variant})") as tq:
                 for i in range(0, B, chunk):
                     end = min(i + chunk, B)
-                    res.extend(face_landmarker.detect_batch(
+                    res.extend(face_detection_model.detect_batch(
                         [img_np[bi] for bi in range(i, end)],
                         num_faces=int(num_faces),
                         score_thresh=float(min_confidence),
@@ -306,7 +309,7 @@ class MediaPipeFaceLandmarker(io.ComfyNode):
                 per_bb.append({"x": x1, "y": y1, "width": x2 - x1, "height": y2 - y1, "label": "face", "score": float(f["score"])})
             bboxes.append(per_bb)
         return io.NodeOutput({"frames": frames, "image_size": (H, W),
-                              "connection_sets": face_landmarker.connection_sets}, bboxes)
+                              "connection_sets": face_detection_model.connection_sets}, bboxes)
 
 
 # Topology keys unioned by the 'all' connections preset (contour parts + irises + nose).
@@ -332,8 +335,10 @@ class MediaPipeFaceMeshVisualize(io.ComfyNode):
     def define_schema(cls):
         return io.Schema(
             node_id="MediaPipeFaceMeshVisualize",
-            display_name="MediaPipe Face Mesh Visualize",
+            search_aliases=["face", "facial", "mediapipe", "face landmark", "face mesh", "blazeface", "face detection", "visualize"],
+            display_name="Visualize Face Landmarks (MediaPipe)",
             category="image/detection",
+            description="Draws face landmarks mesh on the input image.",
             inputs=[
                 FaceLandmarksType.Input("face_landmarks"),
                 io.Image.Input("image", optional=True, tooltip="If not connected, a black canvas will be used."),
@@ -443,8 +448,10 @@ class MediaPipeFaceMask(io.ComfyNode):
     def define_schema(cls):
         return io.Schema(
             node_id="MediaPipeFaceMask",
-            display_name="MediaPipe Face Mask",
+            search_aliases=["face", "facial", "mediapipe", "face mask", "blazeface", "face detection", "visualize"],
+            display_name="Draw Face Mask (MediaPipe)",
             category="image/detection",
+            description="Draws a mask from face landmarks.",
             inputs=[
                 FaceLandmarksType.Input("face_landmarks"),
                 io.DynamicCombo.Input(

comfy_extras/nodes_moge.py

@@ -103,8 +103,10 @@ class MoGePanoramaInference(io.ComfyNode):
     def define_schema(cls):
         return io.Schema(
             node_id="MoGePanoramaInference",
-            display_name="MoGe Panorama Inference",
+            search_aliases=["moge", "panorama", "depth", "geometry", "depth estimation", "geometry estimation"],
+            display_name="Run MoGe Panorama Inference",
             category="image/geometry_estimation",
+            description="Run MoGe on an equirectangular panorama by splitting it into 12 perspective views, running inference on each, and merging the results into a single depth map.",
             inputs=[
                 MoGeModelType.Input("moge_model"),
                 io.Image.Input("image", tooltip="Equirectangular panorama (any aspect)."),
@@ -222,7 +224,9 @@ class MoGeInference(io.ComfyNode):
     def define_schema(cls):
         return io.Schema(
             node_id="MoGeInference",
-            display_name="MoGe Inference",
+            search_aliases=["moge", "depth", "geometry", "depth estimation", "geometry estimation"],
+            display_name="Run MoGe Inference",
+            description="Run MoGe on a single image to estimate depth and geometry.",
             category="image/geometry_estimation",
             inputs=[
                 MoGeModelType.Input("moge_model"),
@@ -277,7 +281,9 @@ class MoGeRender(io.ComfyNode):
     def define_schema(cls):
         return io.Schema(
             node_id="MoGeRender",
-            display_name="MoGe Render",
+            search_aliases=["moge", "render", "geometry", "depth", "normal"],
+            display_name="Render MoGe Geometry",
+            description="Render a depth map or normal map from geometry data",
             category="image/geometry_estimation",
             inputs=[
                 MoGeGeometry.Input("moge_geometry"),
@@ -342,7 +348,9 @@ class MoGePointMapToMesh(io.ComfyNode):
     def define_schema(cls):
         return io.Schema(
             node_id="MoGePointMapToMesh",
-            display_name="MoGe Point Map to Mesh",
+            search_aliases=["moge", "mesh", "geometry", "point map"],
+            display_name="Convert MoGe Point Map to Mesh",
+            description="Convert a MoGe point map into a 3D mesh.",
             category="image/geometry_estimation",
             inputs=[
                 MoGeGeometry.Input("moge_geometry"),

folder_paths.py

@@ -60,7 +60,7 @@ folder_names_and_paths["geometry_estimation"] = ([os.path.join(models_dir, "geom
 
 folder_names_and_paths["optical_flow"] = ([os.path.join(models_dir, "optical_flow")], supported_pt_extensions)
 
-folder_names_and_paths["mediapipe"] = ([os.path.join(models_dir, "mediapipe")], supported_pt_extensions)
+folder_names_and_paths["detection"] = ([os.path.join(models_dir, "detection")], supported_pt_extensions)
 
 output_directory = os.path.join(base_path, "output")
 temp_directory = os.path.join(base_path, "temp")

openapi.yaml

@@ -1556,12 +1556,6 @@ paths:
             type: string
             enum: [asc, desc]
           description: Sort direction
-        - name: job_ids
-          in: query
-          schema:
-            type: string
-          x-runtime: [cloud]
-          description: "[cloud-only] Comma-separated UUIDs to filter assets by associated job."
         - name: include_public
           in: query
           schema:
@@ -2514,37 +2508,25 @@ paths:
 
   /api/assets/import:
     post:
-      operationId: importAssets
+      operationId: importPublishedAssets
       tags: [assets]
-      summary: Import assets from external URLs
-      description: "[cloud-only] Imports one or more assets from external URLs into the cloud asset store."
+      summary: "[cloud-only] Import published assets into the caller's library"
+      description: |
+        [cloud-only] Imports the specified published assets into the caller's asset library. New DB records reference the same storage objects; no file copying occurs. Assets the caller already owns (by hash) are deduplicated. The `id` field on each returned `AssetInfo` is the caller's newly-created private asset ID, not the published asset ID supplied in the request.
       x-runtime: [cloud]
       requestBody:
         required: true
         content:
           application/json:
             schema:
-              type: object
-              required:
-                - imports
-              properties:
-                imports:
-                  type: array
-                  items:
-                    $ref: "#/components/schemas/AssetImportRequest"
-                  description: Assets to import
+              $ref: "#/components/schemas/ImportPublishedAssetsRequest"
       responses:
         "200":
-          description: Import initiated
+          description: Successfully imported assets
           content:
             application/json:
               schema:
-                type: object
-                properties:
-                  assets:
-                    type: array
-                    items:
-                      $ref: "#/components/schemas/Asset"
+                $ref: "#/components/schemas/ImportPublishedAssetsResponse"
         "400":
           description: Bad request
           content:
@@ -7379,24 +7361,35 @@ components:
           type: string
           description: Target path on the runtime filesystem
 
-    AssetImportRequest:
+    ImportPublishedAssetsRequest:
       type: object
       x-runtime: [cloud]
-      description: "[cloud-only] A single asset to import from an external URL."
+      description: "[cloud-only] Request body for importing published assets into the caller's library."
       required:
-        - url
+        - published_asset_ids
       properties:
-        url:
-          type: string
-          format: uri
-          description: URL of the asset to import
-        name:
-          type: string
-          description: Display name for the imported asset
-        tags:
+        published_asset_ids:
           type: array
+          description: IDs of published assets (inputs and models) to import.
           items:
             type: string
+        share_id:
+          type: string
+          nullable: true
+          description: |
+            Optional. Share ID of the published workflow these assets belong to. When provided (non-null, non-empty): all `published_asset_ids` must belong to this share's workflow version; returns 400 if the share is not found or any asset does not belong to it. When omitted, null, or empty string: no share-scoped validation is performed and the assets are validated only against global rules (preserved for clients that have not yet adopted `share_id`).
+
+    ImportPublishedAssetsResponse:
+      type: object
+      x-runtime: [cloud]
+      description: "[cloud-only] Response after importing published assets. Each returned `AssetInfo.id` is the caller's newly-created private asset ID, not the published asset ID supplied in the request."
+      required:
+        - assets
+      properties:
+        assets:
+          type: array
+          items:
+            $ref: "#/components/schemas/AssetInfo"
 
     RemoteAssetMetadata:
       type: object