Full ComfyUI-KJNodes coverage (~200 nodes)
Previously only the ~14 LTXV-subset KJNodes were classified. This
sweep covers the entire NODE_CONFIG dict in
comfyui-kjnodes/__init__.py:
- Loaders: CheckpointLoaderKJ, DiffusionModelLoaderKJ, VAELoaderKJ,
GGUFLoaderKJ, DiffusionModelSelector, every block-lora-loader
variant (Flux / Hunyuan / Wan21 / DiT), IntrinsicLoraSampling,
DownloadAndLoadCLIPSeg, LoadResAdapterNormalization.
- Samplers: GenerateNoise, FlipSigmasAdjusted, InjectNoiseToLatent,
CustomSigmas, FloatToSigmas, SigmasToFloat,
StableZero123_BatchSchedule, SV3D_BatchSchedule.
- Conditioning: every ConditioningSetMaskAndCombine* variant,
WanImageToVideoSVIPro, StyleModelApplyAdvanced,
GLIGENTextBoxApplyBatchCoords, ScheduledCFGGuidance,
ApplyRifleXRoPE * (Hunuyan/Wan), WanVideoTeaCacheKJ,
WanVideoEnhanceAVideoKJ, SkipLayerGuidanceWanVideo, WanVideoNAG,
HunyuanVideoEncodeKeyframesToCond, CFGZeroStarAndInit,
ModelPatchTorchSettings, PatchModelPatcherOrder, LatentInpaintTTM,
NABLA_AttentionKJ, LeapfusionHunyuanI2VPatcher,
DifferentialDiffusionAdvanced, LTXVEnhanceAVideoKJ, all 3
InstanceDiffusionTracking nodes, every TorchCompile* model patch
(FluxAdvancedV2, VAE, ControlNet, WanVideoV2, ModelAdvanced, plus
the deprecated LTX/Cosmos/HyVideo/QwenImage/WanVideo variants),
PathchSageAttentionKJ, CheckpointPerturbWeights,
CustomControlNetWeightsFluxFromList,
SetShakkerLabsUnionControlNetType.
- Decoders: VAEDecodeLoopKJ.
- Outputs / preview / debug: SaveImageWithAlpha, SaveImageKJ,
SaveStringKJ, ModelSaveKJ, LoraExtractKJ, LoraReduceRank,
FastPreview, PreviewAnimation, PreviewLatentNoiseMask,
ImageAndMaskPreview, VisualizeSigmasKJ, CameraPoseVisualizer,
BboxVisualize, VRAM_Debug, TimerNodeKJ, the full CUDA memory
history trio.
- Image-input: LoadAndResizeImage, LoadImagesFromFolderKJ,
LoadVideosFromFolder, ImageGrabPIL, Screencap_mss,
WebcamCaptureCV2, EmptyLatentImagePresets / Custom, ImagePass,
ModelPassThrough, BOOL/INT/FloatConstant.
- Prompts: StringConstant(Multiline), JoinStrings(Multi),
AppendStringsToList, WidgetToString, SomethingToString,
StringToFloatList, ScaleBatchPromptSchedule, Superprompt.
- Post (~120 nodes): every ImageBatch*, Image[Pad/Crop/Concat/
Grid/Insert/Replace/Reverse/Shuffle/Split/Transition]*, every
Mask[Batch/Round/Resize/Offset/Remap], Create*Mask* family,
BatchCropFromMask*, BatchUncrop*, SplineEditor, PointsEditor,
audio→mask/float family, ColorMatch, AddLabel, etc.
README updated with a paragraph per pack listing the bucket
assignments.
shootthesound
a30e71e44a