ComfyUI-KJNodes

AddLabel

Adds a text label above or below an image batch, expanding the image height to make room for the label. Fonts are loaded from the pack’s fonts directory.

KJNodes/Text · 2 inputs · 9 parameters · 1 output

AddNoiseToTrackPath

Adds seeded Gaussian noise to tracking coordinates while preserving their visibility data. Separate controls scale horizontal, vertical, and frame-to-frame variation, including optional temporal smoothing.

AppendInstanceDiffusionTracking

Appends another object’s prompt, coordinates, and tracking data to an existing InstanceDiffusion tracking bundle.

KJNodes/InstanceDiffusion · 2 outputs

AppendStringsToList

Converts each of two string inputs to a list when needed, then concatenates them into one ordered string list.

KJNodes/Text · 2 inputs · 0 parameters · 1 output

ApplyRifleXRoPE_HunuyanVideo

Extends the potential frame count of HunyuanVideo using this method: https://github.com/thu-ml/RIFLEx

KJNodes/Hunyuanvideo · 2 inputs · 1 parameter · 1 output

ApplyRifleXRoPE_WanVideo

Extends the potential frame count of HunyuanVideo using this method: https://github.com/thu-ml/RIFLEx

KJNodes/Wan · 2 inputs · 1 parameter · 1 output

AudioConcatenate

Concatenates two audio waveforms in left-to-right or right-to-left order after confirming that their sample rates match.

KJNodes/Audio · 2 inputs · 1 parameter · 1 output

BatchCLIPSeg

Uses CLIPSeg text prompting to segment one image or a batch into masks, with configurable thresholding, blur, and per-batch processing.

KJNodes/Masking · 3 inputs · 8 parameters · 2 outputs

BatchCropFromMask

Finds mask bounds across an image batch, smooths the crop size and center between frames, and extracts uniformly sized corresponding image crops. It also returns the original images, bounding boxes, and crop dimensions for later uncropping.

KJNodes/Masking · 2 inputs · 2 parameters · 5 outputs

BatchCropFromMaskAdvanced

Creates square, frame-smoothed crops of both images and masks from their nonzero mask bounds. It also returns combined crop regions, per-frame and combined bounding boxes, and dimensions for advanced uncropping.

KJNodes/Masking · 2 inputs · 2 parameters · 9 outputs

BatchUncrop

Resizes cropped images back into their recorded bounding boxes and composites them over the original batch. Crop scale, edge selection, and blurred border blending control how each patch is restored.

KJNodes/Masking · 3 inputs · 6 parameters · 1 output

BatchUncropAdvanced

Composites cropped images back into their originals using per-crop or combined masks and recorded bounding boxes. It supports crop rescaling, soft border blending, and either square or source-mask blending.

KJNodes/Masking · 6 inputs · 4 parameters · 1 output

BBOXToBoundingBoxKJ

Compact description unavailable for this node.

BboxToInt

Selects one bounding box from a list and returns its x, y, width, and height as separate integers.

KJNodes/Masking · 1 input · 1 parameter · 6 outputs

BboxVisualize

Draws a selected bounding box over an image so its recorded crop region can be inspected.

KJNodes/Masking · 2 inputs · 2 parameters · 1 output

BlockifyMask

Creates a block mask by dividing the bounding box of each mask into blocks of the specified size and filling in blocks that contain any part of the original mask.

KJNodes/Masking · 1 input · 2 parameters · 1 output

BOOLConstant

Provides a configurable boolean value and returns it unchanged for use elsewhere in the workflow.

KJNodes/Constants · 0 inputs · 1 parameter · 1 output

CameraPoseVisualizer

Renders camera trajectories as a three-dimensional plot from AnimateDiff-Evolved CameraCtrl poses or RealEstate-format camera coordinates.

KJNodes/Misc · 1 input · 7 parameters · 1 output

CFGZeroStarAndInit

https://github.com/WeichenFan/CFG-Zero-star

KJNodes/Experimental · 1 input · 2 parameters · 1 output

CheckpointLoaderKJ

Experimental node for patching torch.nn.Linear with CublasLinear.

KJNodes/Model Loaders · 3 outputs

CheckpointPerturbWeights

Clones a diffusion model and adds seeded Gaussian perturbations to its weights, scaled by each tensor’s standard deviation. Separate multipliers control joint blocks, the final layer, and all remaining blocks.

KJNodes/Experimental · 1 input · 4 parameters · 1 output

ColorMatch

Transfers the color distribution of a reference image to a target image using a selected color-matching method and blend strength. Batch processing and image inversion options are supported.

KJNodes/Image · 2 inputs · 3 parameters · 1 output

ColorMatchV2

Transfers color from a reference batch to a target batch using MKL, histogram, Reinhard, MVGD, or compound methods. It adds a GPU-accelerated Lab-space Reinhard mode and supports adjustable strength and multithreaded CPU processing.

ColorToMask

Creates a mask from pixels whose RGB values fall within a selected distance of a target color. Batched inputs can be processed in configurable sub-batches.

KJNodes/Masking · 1 output

ConditioningMultiCombine

Combines a configurable number of conditioning inputs into one conditioning sequence.

KJNodes/Masking/Conditioning · 2 outputs

ConditioningSetMaskAndCombine

Applies masks, strengths, and set-area behavior to two conditioning inputs, then combines them using the same operations as ComfyUI’s native conditioning-mask and combine nodes.

KJNodes/Masking/Conditioning · 6 inputs · 3 parameters · 2 outputs

ConditioningSetMaskAndCombine3

Applies masks and strengths to three conditioning inputs and combines the resulting conditioning in one node.

KJNodes/Masking/Conditioning · 9 inputs · 4 parameters · 2 outputs

ConditioningSetMaskAndCombine4

Applies masks and strengths to four conditioning inputs and combines the resulting conditioning in one node.

KJNodes/Masking/Conditioning · 12 inputs · 5 parameters · 2 outputs

ConditioningSetMaskAndCombine5

Applies masks and strengths to five conditioning inputs and combines the resulting conditioning in one node.

KJNodes/Masking/Conditioning · 15 inputs · 6 parameters · 2 outputs

CondPassThrough

Passes positive and negative conditioning through unchanged, providing a routing point that can be bypassed even where a Set node cannot accept bypassed inputs.

KJNodes/Misc · 2 outputs

ConsolidateMasksKJ

Consolidates a batch of separate masks by finding the largest group of masks that fit inside a tile of the given width and height (including the padding), and repeating until no more masks can be combined.

KJNodes/Masking · 1 input · 3 parameters · 1 output

ContextWindowsVisualizerKJ

Interactive visualization of ComfyUI context windows: window placement, overlap, fuse weights and per-frame blending, updated live. Optionally applies the settings to a model like the core Context Windows (Manual) node.

1 input · 13 parameters

CreateAudioMask

Deprecated node that reads an audio file and creates a frame sequence of centered circular masks whose radius follows the audio spectrum level. The sequence can be scaled or inverted.

KJNodes/Deprecated · 1 output

CreateFadeMask

Deprecated node that creates a batch of uniform masks whose intensity moves from a start level through a midpoint to an end level. Linear and eased interpolation modes are supported.

KJNodes/Deprecated · 1 output

CreateFadeMaskAdvanced

Creates a mask batch by interpolating scheduled frame/value keyframes with a selected easing curve. Its schedule syntax is compatible with Fizz BatchValueSchedule.

KJNodes/Masking/Generate · 1 output

CreateFluidMask

Runs a two-dimensional fluid simulation and converts its evolving curl and dye fields into an animated image and mask batch. Inflow count, velocity, radius, padding, duration, and inversion are configurable.

KJNodes/Masking/Generate · 2 outputs

CreateGradientFromCoords

Creates a grayscale gradient image by interpolating values along supplied coordinates, with configurable dimensions and interpolation behavior.

KJNodes/Image · 1 output

CreateGradientMask

Creates an animated horizontal gradient mask batch and offsets the gradient progressively across frames. The resulting masks can be inverted.

KJNodes/Masking/Generate · 1 output

CreateInstanceDiffusionTracking

Creates a new InstanceDiffusion tracking bundle from an object prompt and per-frame coordinates, including class identity and frame dimensions.

KJNodes/InstanceDiffusion · 6 outputs

CreateMagicMask

Generates an animated procedural “magic texture” mask by interpolating seeded coordinate transforms across one or more transitions. It returns both the texture mask and its inverse.

KJNodes/Masking/Generate · 2 outputs

CreateShapeImageOnPath

Draws a selected shape at each coordinate along a path to produce an image batch. Shape size, colors, background, and optional growth over frames are configurable.

KJNodes/Image · 2 outputs

CreateShapeMask

Draws circles, squares, or triangles into a mask batch at a specified center. The shape can grow or shrink on each successive frame, and both normal and inverted masks are returned.

KJNodes/Masking/Generate · 2 outputs

CreateShapeMaskOnPath

Draws a selected shape at each coordinate along a path to produce a mask batch, with configurable frame and shape dimensions.

KJNodes/Masking/Generate · 2 outputs

CreateTextMask

Renders text into an image and mask using fonts from the pack’s fonts directory. Different start and end rotations produce an animated text sequence.

KJNodes/Text · 2 outputs

CreateTextOnPath

Renders text centered at each coordinate along a path, producing a mask batch with configurable font, size, rotation, and frame dimensions.

KJNodes/Masking/Generate · 3 outputs

CreateVoronoiMask

Creates an animated Voronoi-line mask by interpolating randomly generated points over time. Point count, line width, motion speed, and output dimensions are configurable, and both normal and inverted masks are returned.

KJNodes/Masking/Generate · 2 outputs

CrossFadeImages

Crossfades a selected segment of one image batch into the beginning of another using configurable start and end blend levels and an easing curve. Frames before the transition and remaining frames from the second batch are retained.

KJNodes/Image · 1 output

CrossFadeImagesMulti

Joins a configurable number of image batches, inserting eased crossfade frames between the last frame of each batch and the first frame of the next. Later inputs are resized to match the first batch when necessary.

KJNodes/Image · 1 output

CustomControlNetWeightsFluxFromList

Creates controlnet weights from a list of floats for Advanced-ControlNet

KJNodes/Controlnet · 3 inputs · 1 parameter · 2 outputs

CustomSigmas

Parses comma-separated sigma values into a sampling schedule and can interpolate that schedule to a requested step count. It also returns a normalized text representation of the final values.

KJNodes/Noise · 0 inputs · 2 parameters · 1 output

CutAndDragOnPath

Cuts a masked region from an image and moves it along a coordinate path to create an animated batch. When inpainting is enabled without a background image, the vacated region is filled with OpenCV’s Telea method.

KJNodes/Image · 2 outputs

DecodeAndSaveVideo

Decodes video frames and audio from latent representations, combines them, and saves as a video file, without keeping intermediate images in memory.

4 inputs · 5 parameters

DifferentialDiffusionAdvanced

Attaches a sampling-step-dependent denoise-mask function to a cloned model and stores the supplied mask as the latent noise mask. The multiplier changes how quickly mask pixels become active over the denoising schedule.

For Testing · 3 inputs · 1 parameter · 2 outputs

DiffusionModelLoaderKJ

Node for patching torch.nn.Linear with CublasLinear.

KJNodes/Model Loaders · 1 output

DiffusionModelSelector

Returns the path to the model as a string.

KJNodes/Model Loaders · 0 inputs · 1 parameter · 1 output

DiTBlockLoraLoader

Loads a LoRA from the model directory or an explicit path and applies it to a cloned diffusion model. An optional block-selection map can remove or rescale LoRA contributions per DiT block, and the node also reports the detected rank.

KJNodes/Lora · 3 inputs · 2 parameters · 2 outputs

DownloadAndLoadCLIPSeg

Downloads the configured CLIPSeg checkpoint from Hugging Face when absent, loads it into ComfyUI, and returns the model for segmentation nodes.

KJNodes/Masking · 0 inputs · 1 parameter · 1 output

DrawInstanceDiffusionTracking

Draws InstanceDiffusion tracking paths and visibility information over an image batch for inspection.

KJNodes/InstanceDiffusion · 1 output

DrawMaskOnImage

Applies the provided masks to the input images with Alpha Blending support.

KJNodes/Masking · 2 inputs · 2 parameters · 1 output

DummyOut

Acts as a generic output node without modifying its input, allowing a workflow to execute and show previews without saving files to disk.

KJNodes/Misc · 1 output

EmptyLatentImageCustomPresets

Generates an empty latent image with the specified dimensions. The choices are loaded from 'custom_dimensions.json' in the nodes folder.

KJNodes/Latents · 0 inputs · 3 parameters · 3 outputs

EmptyLatentImagePresets

Creates an empty latent batch using a selected built-in width and height preset. It can swap the preset dimensions and returns the resolved width and height with the latent.

KJNodes/Latents · 3 outputs

EncodeVideoComponents

Extracts video frames, resizes them, and encodes with a VAE directly, avoiding storing the full image tensor.

2 inputs · 5 parameters

EndRecordCUDAMemoryHistory

Records CUDA memory allocation history between start and end, saves to a file that can be analyzed here: https://docs.pytorch.org/memory_viz or with VisualizeCUDAMemoryHistory node

KJNodes/Memory · 2 outputs

FastPreview

Fast image preview using binary websocket, bypassing base64/JSON overhead.

KJNodes/Experimental · 1 input · 2 parameters · 0 outputs

FastPreviewBatch

Encodes an image batch as an all-I-frame H.264 MP4 thumbnail strip and shows it as an interactive grid. Click a tile to enlarge with prev/next browsing.

FilterZeroMasksAndCorrespondingImages

Removes fully empty masks from a batch and returns their original indices. When images are supplied, it removes the images at the same indices so both batches stay aligned.

KJNodes/Masking · 2 inputs · 0 parameters · 4 outputs

FlipSigmasAdjusted

Reverses a sigma schedule, shifts its entries by a configurable offset, and scales the result by a divisor. It can additionally normalize against the last adjusted sigma and returns a readable sigma string.

KJNodes/Noise · 1 input · 3 parameters · 2 outputs

FloatConstant

Provides a configurable floating-point value, rounded to six decimal places, for use elsewhere in the workflow.

KJNodes/Constants · 0 inputs · 1 parameter · 1 output

FloatToMask

Converts a list of floating-point values into a batch of uniformly valued masks at the requested width and height.

KJNodes/Masking/Generate · 1 input · 2 parameters · 1 output

FloatToSigmas

Converts a list of floating-point values into a SIGMAS tensor for use as a sampling schedule.

KJNodes/Noise · 1 input · 0 parameters · 1 output

FluxBlockLoraSelect

Select individual block alpha values, value of 0 removes the block altogether

KJNodes/Experimental · 0 inputs · 0 parameters · 1 output

GenerateNoise

Generates seeded latent noise at a requested shape for direct injection or for samplers whose built-in noise addition is disabled.

KJNodes/Noise · 2 inputs · 9 parameters · 1 output

GetImageRangeFromBatch

Returns a requested frame range from an image batch, a mask batch, or both while keeping the two outputs aligned.

KJNodes/Image · 2 outputs

GetImagesFromBatchIndexed

Selects and returns the images at the specified indices as an image batch.

KJNodes/Image · 1 output

GetImageSizeAndCount

Returns width, height and batch size of the image, and passes it through unchanged.

KJNodes/Image · 1 input · 0 parameters · 4 outputs

GetLatentRangeFromBatch

Returns a range of latents from a batch.

KJNodes/Latents · 1 output

GetLatentsFromBatchIndexed

Selects specified indices from a latent batch and returns them as a new latent batch.

KJNodes/Latents · 1 output

GetLatentSizeAndCount

Returns latent tensor dimensions, and passes the latent through unchanged.

KJNodes/Image · 1 input · 0 parameters · 6 outputs

GetMaskSizeAndCount

Returns the width, height and batch size of the mask, and passes it through unchanged.

KJNodes/Masking · 1 input · 0 parameters · 4 outputs

GetPreviewOverrideFramesKJ

Returns the frames captured by Model Preview Override during the most recent sampling. Wire 'model' from Model Preview Override (the same one feeding the sampler) and 'after_sample' from after the sampler (LATENT/IMAGE) to enforce correct execution order.

GetTrackRange

Slices tracking paths and their visibility data to a requested frame range. Negative start indices count backward from the end, and the range is clamped to available frames.

GGUFLoaderKJ

Loads a diffusion model through ComfyUI-GGUF with selectable dequantization and patch dtypes, patch-on-device behavior, and attention backend. It can merge an additional GGUF or LTX2 connector model and optionally enable FP16 accumulation.

0 inputs · 7 parameters

GLIGENTextBoxApplyBatchCoords

Applies scheduled text-box coordinates to GLIGEN conditioning across a latent batch. It encodes the supplied text, attaches per-frame bounding boxes, and returns both conditioning and a coordinate preview.

KJNodes/Experimental · 6 inputs · 3 parameters · 2 outputs

GradientToFloat

Samples a grayscale gradient image at supplied coordinates and returns the sampled values as a float list.

KJNodes/Image · 2 outputs

GrowMaskWithBlur

Expands or contracts masks with optional per-frame growth, tapered corners, inversion, blur, interpolation decay, and hole filling.

KJNodes/Masking · 1 input · 8 parameters · 2 outputs

HDRPreviewKJ

Provides a browser-side realtime exposure and saturation preview for LogC3-compressed HDR image or video batches. It also returns a baked sRGB image using the same decompression, tonemapping, and color math.

1 input · 4 parameters

HunyuanVideoBlockLoraSelect

Select individual block alpha values, value of 0 removes the block altogether

KJNodes/Hunyuanvideo · 0 inputs · 0 parameters · 1 output

HunyuanVideoEncodeKeyframesToCond

Builds a HunyuanVideo latent sequence with supplied start and end images, encodes it through the VAE, and attaches the encoded keyframes to positive and negative conditioning. Tiled spatial and temporal encoding controls reduce memory use.

KJNodes/Hunyuanvideo · 6 inputs · 5 parameters · 4 outputs

Ideogram4OptimizationsKJ

Experimentally reduces Ideogram 4 peak VRAM by chunking feed-forward activations across tokens and optionally applying rotary embeddings in the model dtype instead of FP32. Either optimization can change numerical output.

1 input · 4 parameters

Ideogram4PromptBuilderKJ

Provides a visual canvas for assembling Ideogram 4 structured JSON captions with object and text regions, bounding boxes, descriptions, palettes, background, and style fields. It outputs the finished caption JSON string.

ImageAddMulti

Adds a configurable number of image tensors together, using adjustable input slots to build one blended image output.

KJNodes/Image · 1 output

ImageAndMaskPreview

Previews an image, a mask, or a mask composited over an image with configurable color and alpha. Pass-through mode disables the preview and returns the composite for downstream use.

KJNodes/Masking · 2 inputs · 3 parameters · 1 output

ImageBatchExtendWithOverlap

Prepares or joins frame batches for video extension by selecting an overlap from a source batch and blending or concatenating newly generated frames according to the chosen overlap mode.

KJNodes/Image · 3 outputs

ImageBatchFilter

Removes empty images from a batch

KJNodes/Image · 2 outputs

ImageBatchJoinWithTransition

Transitions between two batches of images, starting at a specified index in the first batch. During the transition, frames from both batches are blended frame-by-frame, so the video keeps playing.

KJNodes/Image · 2 inputs · 7 parameters · 1 output

ImageBatchMulti

Combines a configurable number of image inputs into one image batch. Inputs that differ from the first image's dimensions are resized to match, and channel counts are padded before the frames are concatenated.

KJNodes/Image · 1 output

ImageBatchRepeatInterleaving

Repeats each individual image in a batch consecutively by the requested count, preserving the original frame order. An optional mask batch is interleaved in the same way.

KJNodes/Image · 2 inputs · 1 parameter · 2 outputs

ImageBatchTestPattern

Generates an image batch containing sequential frame numbers rendered with a chosen font, colors, dimensions, and starting index.

0 inputs · 8 parameters

ImageConcanate

Concatenates image2 to image1 in the specified direction. Both inputs accept IMAGE or MASK; the output type follows image1.