🔧 Image Color Match Adobe

When a composite or an upscale comes out with the wrong color cast - the pasted subject is too warm, the refined region drifted cooler than the rest - you need to make one image match another's palette. Image Color Match Adobe does that. It takes your image plus a reference and shifts your image's colors toward the reference's. The "Adobe" in the name is the tell: it's modeled on Photoshop's classic Match Color command, so if you've used that, you already know the effect.

What it does

Two images in, one out. Your image gets recolored to resemble the reference. The controls are what make it usable rather than a blunt instrument:

• color_space - RGB or LAB. LAB separates lightness from color, which usually gives a more natural match because it can shift hue without wrecking brightness. RGB is more literal. LAB is the better default for most photographic work.
• luminance_factor (0–2, default 1) - how strongly to match brightness/lightness. Turn it down if you want the reference's color but not its exposure.
• color_intensity_factor (0–2, default 1) - how strongly to match saturation/chroma. Down for a subtle tint, up to really force the palette.
• fade_factor (0–1, default 1) - the overall blend between your original and the fully-matched result. This is your master "how much" dial; 1 is full match, 0.5 is halfway. When a full match looks overcooked, this is the knob you back off.
• neutralization_factor (0–1, default 0) - pulls color casts toward neutral, useful for removing an unwanted tint rather than adopting a new one.
• device - auto, cpu, or gpu. Leave on auto unless you're chasing a specific performance or memory issue.

There's also an optional reference_mask so you can match against only part of the reference - sample the sky's color, not the whole frame, for instance.

Output is the recolored IMAGE.

Where it fits

Two big jobs. Compositing: you pasted a subject onto a background and it doesn't sit right - match the subject to the background and the seam vanishes. Consistency across a batch or a pipeline: a hires-fix or a second-model pass shifted the palette, so match the output back to the original to keep a series looking coherent. It's also handy for grading a whole set of images to one hero frame's look.

Installing it

Part of ComfyUI Essentials by cubiq - Matteo Spinelli, the developer behind ComfyUI_IPAdapter_plus:

• ComfyUI Manager: search "ComfyUI Essentials" → Install → restart.
• Manual: cd ComfyUI/custom_nodes && git clone https://github.com/cubiq/ComfyUI_essentials, install requirements.txt, restart.

Gotchas

The default full match (all factors at 1, fade 1) is often too aggressive - it'll happily make your image look like a bad filter. Pull fade_factor down to taste; a 40–70% match usually reads more natural than a full one.

Your reference choice is everything. Match against an image whose palette you actually want, and if only part of it matters, use reference_mask to sample just that region - otherwise a busy reference averages into a muddy target.

LAB vs RGB genuinely changes the result; if RGB is giving you weird brightness shifts, switch to LAB and it usually calms down.

Pack note: cubiq set Essentials to maintenance-only in April 2025. It still works, but if the node errors after a big ComfyUI update, roll ComfyUI back or patch it - no upstream fixes. A red "missing node" box means the pack didn't load; check the startup log and reinstall via Manager.