LC Image RGB
The three-knob color cast fixer you didn't know you needed
- image
- image
Your render came out green. Or orange. Or looks like someone forgot to turn the white balance on. Fixing it doesn't need a diffusion pass, an img2img reroll, or anything generative - it needs a per-channel offset, which is exactly what LC Image RGB is. Three sliders, one per channel, and the whole frame shifts. It's the most boring grading tool in this pack and one of the most useful.
How it works
The node adds a value to each RGB channel and clamps the result to 0–1. Each of r, g, b runs −1 to 1 with 0.01 steps. Negative pulls a channel down (drains green), positive pushes it up (adds red). Since it's additive - add, not multiply - it moves shadows and highlights by the same amount, which is the correct tool for a cast that's uniform across the frame. A strength slider (0–1, default 1) blends the offset result back toward the original, so you can dial a correction in gently instead of committing fully.
The outputs are the same inputs, which sounds silly until you chain: this is one of a family of FX nodes in the pack that show a before/after wipe when you hover the node, so you're grading against the original without dragging extra preview nodes around. Output is image, ready for the next stage.
The knobs that matter
Honestly, only three of them do anything on their own:
r- shift the red channel (negative cools, positive warms).g- the one people forget; a +0.05 green nudge fixes a magenta cast that no WB slider will touch.b- negative drags yellows out of skin; positive adds a cool blue.
strength is the one to ride while you're dialing in. A common sequence: push the channel you suspect, check the wipe, then back off with strength until the cast is gone and the image still has life.
Install
Part of ComfyUI LC123 Nodes (MIT, 89 nodes). ComfyUI Manager → "ComfyUI LC123 Nodes", or:
cd ComfyUI/custom_nodes
git clone https://github.com/lonecatone23/ComfyUI_LC123_nodes
Restart after. No extra pip packages, no model downloads. On cloud hosts like RunningHub, the platform installs the pack for you on request.
When not to use it
A color cast that lives only in the shadows isn't an offset job - that's a lift/gamma/gain problem (LC Lift Gamma Gain in the same pack does shadows/mids/highlights separately). And if the image is "flat" rather than tinted, you want a gamma or contrast curve, not per-channel math. This is the scalpel for casts; reach for the right primitive and you'll stop reaching for img2img entirely.
Inputs (5)
| Name | Type | Default | Description |
|---|---|---|---|
| image | IMAGE | — | |
| r | FLOAT | 0.00-1–1 | Red channel offset (-1..1) |
| g | FLOAT | 0.00-1–1 | Green channel offset (-1..1) |
| b | FLOAT | 0.00-1–1 | Blue channel offset (-1..1) |
| strength | FLOAT | 1.000–1 | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| image | IMAGE | — |