ComfyUI-YCNodes
A collection of image processing extension nodes for ComfyUI.
Nodes (57)
Rotate through 10 images — randomly, cyclically, or by index
A fixed lookup table that maps an input number to a CFG — know what you're getting
Pad an image to a target ratio — and get the mask that proves what's original
Pull one named image out of a batch — without counting frames
Drop a foreground onto a background by telling it WHERE — a mask does the locating
Place, scale, rotate, and blend one image onto another — with the mask to show it
Stitch two images into one canvas for in-context LoRA work
The in-context combiner with pixel-level control — overlay, XY placement, and the main image on its own
Load an image and its alpha mask in one shot — blur and strength built in
Mirror an image — yes, it's really that simple
Pixelate a face (or anything) without a trip to Photoshop
Scale an image and its mask as one unit — then hand the original size back so you can restore it
Pick the keeper out of an 8-image batch — stop upscaling the other seven
Pick a latent size that actually matches your image's aspect ratio
Tile upscale without nuking your VRAM
Turn a messy hand-drawn mask into a clean, regular shape
Load a whole folder of images — single, batch, random, or stepping through one by one
Staple two mask batches together — despite the name, it's concatenation, not blending
Duplicate a mask batch into two outputs — the splitter you didn't know you needed
Stamp one mask onto another at any position — with multiply, XOR, overlay, and friends
Fill in a line-art or outline mask so it becomes a solid region
Stitch a cropped, edited image back onto its background — the 'restore' half of crop-edit-stitch
Crop exactly around a mask — padding, multiple-of-8 sizing, and a live preview of the box
Keep the solid blobs, drop the fragments — filter masks by how convex they are
The mask slicer ComfyUI core forgot
Actually look at your mask — a preview node that doesn't fake it with black boxes
One mask, a batch of frames — repeat it to match
Pad your mask's content out to a clean aspect ratio — no canvas resize involved
Split a joined mask along its waist — where two blobs are fused, cut the valley
Explode a multi-object mask into one mask per object
Keep the N biggest blobs in a mask, delete the rest
Strip (parenthetical noise) out of your prompt — full-width brackets included
Grab everything before (or after) a keyword — for filenames, tags, and prompt surgery
Fit any batch of images inside max width/height — and know it's a list coming out
Pull the leading number out of a string — the tiny logic helper that saves a regex node
Five text boxes, one node — write your prompt in parts and get the join for free
Batch up every generation, then pick the winner after
A save node that actually names your files and writes captions too
Crop an image and get the map of where you cropped it
Pad a canvas and get a mask of the new border — your outpainting starts here
Split a big image into overlapping tiles — the cheap ticket to upscaling past VRAM
One image, repeated into a contact sheet — the node for 'show me all the variants' boards
Stitch tiles back into a full image — the other half of tiled upscaling
Feather your masks before they hit the sampler — the difference between a seam and a blend
Paint your mask onto the image — the fastest way to see (and abuse) what you're masking
Route between two masks based on a text match — a tiny switch for loop workflows
Grow a mask in one direction without wrecking the other
Center your subject in a ratio-perfect frame — pad, crop, and hand back every mask you need
Stretch a dim mask up (or pull a hot mask down) — simple range surgery for masks
Give every iteration of a loop its own seed — deterministic batch variety
Just a text box — the node you didn't know you wanted
A ten-slot LoRA switcher you can chain — and it only loads the LoRA you picked
Pick between two images based on a text match — the loop workflow's cheap branch
Render text to an image (and its mask) without ever leaving ComfyUI
Pick one of 12 prompts with an index — and only compute that one
Find-and-replace for your prompt, up to three swaps deep — dead simple, occasionally perfect
A pass-through gate for ANY data type, opened by a text condition — including a not-switch
ComfyUI-YCNodes
A collection of image processing extension nodes for ComfyUI.
Nodes Description
Image mirror, image rotate, image mosaic
ImageSelector
Multiple input images, select by custom name, select the specified image output (Ten fixed inputs)
ImageBatchSelector
Image combination batch, receiving batch image input, and selecting output specified images by custom name. Theoretically, there is no limit to the number of inputs.
Image IC
Basic image combination node with the following features:
- Horizontal/vertical image concatenation
- Aspect ratio scaling based on reference edges
- Flexible position adjustment (top/center/bottom/left/right)
- Custom background color
- Mask processing support
Image IC Advanced
Advanced image combination node, adding the following features to the basic version:
- Overlay mode support
- Precise XY coordinate control
- Independent scaling control
- Additional output options
This two node provides advanced image combination functions, designed for IClora and fill+redux partial redraw migration images. Optimize the combination of two graphs, save the size of Flux iclora and Fill+redux when partial redrawing, and save computing power.
The basic code of the nodes ic image and ic image advanced comes from https://github.com/lrzjason/Comfyui-In-Context-Lora-Utils Based on it, I optimized two nodes: ic image and ic image advanced. Thanks to the original author @小志Jason
Installation
Clone this repository into ComfyUI's custom_nodes directory:
cd custom_nodes
git clone https://github.com/your-username/Comfyui-yicheng-node.git
关于我 | About me
Bilibili:我的B站主页 QQ号 3260561522 wechat微信: DLONG189one