Image Blank

Make a solid-color image in ComfyUI

By WASasquatch·Created 3 years ago·Updated 2 days ago· 1,839

Image Blank

IMAGE

◄width512►

◄height512►

◄red255►

◄green255►

◄blue255►

◄divisible_by8►

◄batch_size1►

Sometimes you don't need a generated image, you need a flat rectangle of one color. A solid backdrop to composite onto. A white canvas to paste a crop into. A colored fill to blend against. Image Blank does exactly that and nothing else: pick a size, pick a color, out comes a solid image.

It's the kind of node you don't think about until a workflow suddenly needs one, and then you're glad it exists. Core ComfyUI makes you jump through hoops to get a plain colored image; this hands it to you in one node.

How it works

There's no generation here - no model, no sampling, no noise. It allocates an image at the width and height you give it and fills every pixel with the color you specify. Instant, deterministic, free. Because the output is a standard IMAGE tensor, it drops straight into any node that takes an image: blend nodes, paste/composite nodes, mask operations, or just a preview so you can eyeball the color.

The classic uses: a base layer for compositing something on top, a solid color to blend into an existing image (a tint or wash), a placeholder while you build a graph, or a clean canvas for the paste-crop family of nodes so a crop has somewhere to land.

The inputs and outputs that matter

You set the width and height for the dimensions, and the color as red/green/blue channel values. That's it - three things, and two of them are just "how big." The output is a single image at that size, filled with that color, ready to wire anywhere downstream.

If you want the canvas to match an existing image's size, feed its dimensions in rather than typing numbers - WAS's Image Size to Number pulls width and height off an image so your blank always matches.

How to install it

Ships in WAS Node Suite. The painless way is ComfyUI Manager: search WAS Node Suite, install, restart. Manually:

cd ComfyUI/custom_nodes
git clone https://github.com/WASasquatch/was-node-suite-comfyui
pip install -r was-node-suite-comfyui/requirements.txt

then restart ComfyUI. On the Windows portable build use python_embeded\python.exe -s -m pip install -r ... (or the included install.bat) so the requirements install into ComfyUI's Python rather than your system one. Image Blank itself needs no models or extra libraries - it's about as lightweight as a node gets.

Common issues & troubleshooting

The color isn't what you expected. The color is set per RGB channel (0–255), so 0,0,0 is black, 255,255,255 is white, and mixing gives you everything between. If you get an unexpected shade, it's almost always a channel value typed wrong - double-check the three numbers before blaming the node.

Size mismatch when compositing. Blend and paste nodes generally want the layers to line up. If your blank is 512×512 and the image you're pasting onto is 768×1024, expect trouble. Match the blank to the target - feed real dimensions in instead of hardcoding them, and it stays correct even when the upstream size changes.

It's not the fastest way anymore. This node is fine and it works, but if you have a modern utility pack loaded (something like the newer image-primitive nodes), you may already have a solid-color generator that fits your graph with less clutter. No reason to install all of WAS just for this - but if the suite's already there, Image Blank is the obvious pick.

The suite won't load at all. WAS Node Suite is a big, no-longer-maintained pack, and the recurring gotcha is an "Import Failed" after you update ComfyUI, caused by a dependency version clash (opencv especially). Reinstall the requirements against your ComfyUI Python and restart. The node is trivial; the suite's install is the only real risk.

CategoryWAS Suite/Image/Generate

Inputs (7)

Name	Type	Default	Description
width	INT	5128–4096	Width in pixels, rounded down to a multiple of divisible_by: 500 gives 496 at the default of 8, and 500 at a divisible_by of 1.
height	INT	5128–4096	Height in pixels, rounded down to a multiple of divisible_by: 500 gives 496 at the default of 8, and 500 at a divisible_by of 1.
red	INT	2550–255	Red level of the fill colour. 0 is none, 255 is full.
green	INT	2550–255	Green level of the fill colour. 0 is none, 255 is full.
blue	INT	2550–255	Blue level of the fill colour. 0 is none, 255 is full. All three at 255 gives white, all three at 0 gives black.
divisible_by	INT	81–64	Rounds width and height down to a multiple of this. 8 suits most latent models; set it to 1 to get the exact canvas asked for.
batch_size	INT	11–4096	How many copies the batch holds. 1 = a single image; 16 = sixteen identical fills, for matching a batch of frames a sampler or a video node is working on.

Outputs (1)

Name	Type	Description
IMAGE	IMAGE	A batch of batch_size images, each filled edge to edge with the chosen colour, at the requested size rounded down to a multiple of divisible_by, with a side shorter than that taken up to one whole step instead.