CONSTANT (JOV) πͺ
A solid color, a matte, or a background β on demand
- πΌπ
- π
- πΌοΈ
- π·
CONSTANT (JOV) πͺ makes a solid-color image of a size and color you specify - and then gives you the mask to go with it. In a pack full of flashy compositing tools, this is the quiet utility you'll actually use constantly: a background to composite onto, a color matte to grade against, a resolution-forcing blank, or just a test card while you debug a workflow.
It's the current name for what used to be "Constant (jov)" before Jovimetrix's 2.0 renaming pass - same job, upgraded inputs. The old version used three 0β1 float sliders for R/G/B; this one gives you a VEC4 color with a proper picker.
What it does
Two inputs, two outputs:
- πΌπ (VEC2, default
(512, 512), 32β8192) - the width and height of the output. This is the one you'll set every time. - π (VEC4, default
(0, 0, 0, 255), 0β255) - the color, as RGBA in 0β255 space. The 255 alpha default means fully opaque; drop it for a transparent constant. This is the upgrade over the legacy node - a real color input with picker support.
Outputs are πΌοΈ image and π· mask: the image is your solid color, and the mask is a fully-white matte at the same size. In newer builds the node also accepts an optional input image and mask, letting you matte a chosen color onto an existing image rather than starting from scratch - the README notes "patched constant node to work with MATTE_RESIZE," which is that behavior.
How it works
It allocates a canvas at the requested size, fills it with the color, and generates the corresponding alpha. If you feed it an optional image/mask, it composites the color under that image using the mask - which is how you'd tint or recolor a background without touching the foreground. Pure pixel work, no model, instant.
Installing it
Part of Jovimetrix:
- ComfyUI Manager - search "Jovimetrix", install.
- Manual -
git clone https://github.com/Amorano/Jovimetrix.gitintocustom_nodes/, thenpip install -r requirements.txt.
No model downloads. Dependencies: numpy, OpenCV (opencv-contrib-python), Pillow, matplotlib, plus the author's cozy_comfyui library from GitHub. Needs ComfyUI 0.1.3+ and a reasonably recent frontend.
Where people get burned
Two things, and they're both about color conventions. First, the 0β255 range: if you type 0.5 in the alpha expecting a 50% constant, you get a 0.5/255 - basically invisible. Use 255 for opaque, 128 for half. Second, if your constant comes out fully black when you expected white, check that you changed all four values - the default is (0,0,0,255), black with full alpha. The mask output is worth grabbing even when you don't need it: it's a ready-made full matte at exactly the constant's size, which is a handy sizing anchor for downstream compositing. It's not a node you'll talk about - it's the node that makes the ones you do talk about possible.
Inputs (2)
| Name | Type | Default | Description |
|---|---|---|---|
| πΌπopt | VEC2 | 512,51232β8192 | β |
| πopt | VEC4 | 0,0,0,2550β255 | β |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| πΌοΈ | IMAGE | β |
| π· | MASK | β |