Variable Image Builder (Soze)
A solid RGBA image, made to spec
- Image
A solid-color image generator, like ComfyUI's stock Empty Image node but with a full RGBA channel set instead of just RGB, and explicit control over batch count. Reach for it whenever you need a flat, known image rather than a real photo - a backdrop to composite onto with this pack's own Image Overlay node, a placeholder while you're building out a graph, or a uniform base feeding into a comparison workflow.
How it works
You specify each color channel independently and the node fills a canvas of the given size with it - one image, or a batch of identical ones if you want more than one copy.
The inputs and output that matter
- r, g, b, a - the four 8-bit color channels, 0–255 each. Notably, these don't come with a default value the way
width/height/batch_sizedo - set them explicitly, or you'll likely land on the widget's minimum (0), which gives you solid black at full transparency rather than anything useful. - width / height - canvas size in pixels, both default 512.
- batch_size - how many identical copies to generate in one batch, default 1, minimum 1.
Output is a single Image.
Where it fits
Two concrete uses show up a lot: as a base layer for compositing, where you build a solid backdrop and then paste other elements onto it with something like this pack's Image Overlay node instead of loading a stock background photo every time; and as a stand-in during graph development, where you don't want to wait on a real generation just to check that your downstream nodes (resize, save, overlay) are wired correctly - a flat color image is instant and tells you immediately if the plumbing works. It's also a reasonable partner for this pack's own Pad Mask: build a solid image at the size you're about to pad out to, so you can preview exactly how much new canvas your padding values actually add before you commit to running a full outpaint.
Installing it
ComfyUI Manager: search ComfyUI_Soze. Manual: cd ComfyUI/custom_nodes && git clone https://github.com/SozeInc/ComfyUI_Soze.git, pip install -r ComfyUI_Soze/requirements.txt, restart. No models, no dependencies - it's a pixel fill operation.
Common issues
If your output image comes out solid black and you didn't set r/g/b/a explicitly, that's the missing-default behavior above, not a bug. And keep in mind not every downstream node handles the alpha channel the same way - if something further along the chain expects plain RGB, it may silently drop or mishandle the a value. Set a to 255 unless you specifically need transparency, so you're not debugging a compositing issue that's actually just an unused alpha channel.
Inputs (7)
| Name | Type | Default | Description |
|---|---|---|---|
| r | INT | 0–255 | — |
| g | INT | 0–255 | — |
| b | INT | 0–255 | — |
| a | INT | 0–255 | — |
| width | INT | 512 | — |
| height | INT | 512 | — |
| batch_size | INT | 1 | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| Image | IMAGE | — |