Image Processing Suite for ComfyUI

Noise That Respects Your Edges (a GIMP-Grade Trick)

Ken Burns Your Batch Without OOMing

Turn detection boxes into SAM2 point prompts — the glue between YOLO and SAM

Cut Between Up to 8 Cameras Automatically

Depth Maps Straight Out of the Box, No Preprocessor Pack Needed

ImageCompositeMasked without the OOM — the memory-safe paste node

Image to Mask — pull one channel out of an image and use it as a mask

The Loader That Fetches Depth Anything V3 For You

Load Segment Anything 2 — for stills or video — weights fetched automatically

Load a YOLO detection model, downloaded automatically on first use

Automated LTX keyframing — count and positions come from your index list

Mask Gaussian Blur — the feathering tool for masks that came out too sharp

Add, Subtract, AND and OR Your Masks Like It's Boolean Algebra

Turn a mask into a viewable, savable RGB image

Pick a body part, get the keypoints as SAM2-ready JSON

SAM2 Image Inference — point at something and get a mask, no extra node packs required

Teach SAM2 What to Track on One Frame

Let SAM2 Chase the Subject Through Every Frame

Save images with a text tag baked into the PNG — dataset curation made searchable

The Debug Node That Finally Shows You What's In That List

Smooth Pan and Zoom Keyframes Without the Math

Randomized Ken Burns Motion For Whole Video Batches

Detect People, Poses and Boxes Inside ComfyUI, No Separate Repo

Stitching two audio clips end to end

Grab One Audio Slice Out of a List, Sans Pain

Cut Audio at Exact Timestamps Into Slices You Can Reuse

Split a batch into a list so ComfyUI processes frames one at a time

The caption loader that turns a folder into a training set

Turn ten dropdowns into a full character prompt

A fighting-game OC generator that respects its seed

Perceptual film grain, straight from GIMP's math

Comparison Swipe Video — before/after as an animated wipe, straight out of the node

CSV Prompt Loader — your prompt library in a spreadsheet, selected from a dropdown

Crop an image and its mask in lockstep, so inpainting never misaligns

Dataset Loader — images and their captions, loaded as matching lists for training prep

Fighting Game Character Generator — roll a fighter profile into a prompt

Pull the prompt back out of every image in a folder — your dataset just got easier

Folder Image Loader — point it at a directory and get images, masks, and filenames back

Pull one specific node's text out of every image in a folder

Concatenate every MP4 in a folder — no re-encoding, no tears

Grab frame zero out of a batch

Slice One Image Into a Grid of Tiles (and Feed a Batch)

Evenly pick N frames from a sequence, and get their positions too

Turn a list of images into one batch tensor

Find out whether your IMAGE is a batch or a single frame, and get both

The boring 50-output adapter that unblocks LTX keyframing

Interactive Crop — drag a box on the image, get the crop and its coordinates

Get the Last Image from a batch — boring, tiny, and you'll use it more than you expect

Load a GIF's unique frames and the map to rebuild the full sequence

Load a PNG and get its prompt back out of the metadata

The loader that composites transparency onto white, so sprites stop leaking

Load an entire folder of images at native resolution in one node

When Your LTX Video Latents Look Weird, This Tells You Why

Put N keyframes inside an LTX latent without the attention tax

The smoother sibling of LTX Multi Concat

Inject N reference images into an LTX Video latent at exact frame positions

The vid2vid encoder LTX never shipped in its example workflows

Point SAM2 At the Right Spot Without the JSON Headache

An edge-preserving blur that doesn't wreck the lines

Crisp nearest-neighbor upscaling, made for pixel art

Make Real Pixel Art, Not Just Downscaled JPEGs

Pixel Art Converter — an actual pixel-art algorithm, not a Resize node wearing a costume

Detect the pixel grid and true-up any image to clean pixel art

Click on the image to add SAM2 points — the node that finally makes point prompts usable

Refine Mask — expand or shrink a mask, then feather it, in one node

Reassemble GIF Frames That Other Nodes Sliced Up

Flatten an RGBA image onto white, already inside your graph

GEGL's RGB noise, ported to torch — grain without GIMP

SEGS to BBox — the adapter that lets Impact Pack detections drive SAM2

Turn Impact Pack detections into SAM2 prompt points

Square up any image with border fill that matches its own edges

Browse a Trello board as prompts, right inside ComfyUI

Headless prompt loading from a Trello export

Video + Audio Combiner — mux a soundtrack onto your render without leaving ComfyUI

Mux an MP3 onto your video without touching the pixels

Mix the original audio AND a new track, volumes included

Two Videos, Different FPS, Side by Side Without the Jank

A walk cycle as OpenPose skeletons, procedurally — no model, no ControlNet preprocessing