Nodes/Leon's Utility and API Integration Nodes/πŸ€– Leon Yellow Tint Cleaner
ComfyUI Node

πŸ€– Leon Yellow Tint Cleaner

Fix the yellow cast with zero API calls

By l3ony2kΒ·Created about a year agoΒ·Updated 2 months agoΒ· 3
πŸ€– Leon Yellow Tint Cleaner
  • image
  • image
β—„strength100β–Ί
β—„brightness0β–Ί
β—„contrast0β–Ί
β—„saturation0β–Ί
β—„red0β–Ί
β—„green0β–Ί
β—„blue0β–Ί
β—„modeRGBβ–Ί

Here's the thing nobody tells you about the API pack: two of its nodes are fully local, free, and need no key, and this is the genuinely useful one. Leon_Yellow_Tint_Cleaner_Node is a color-correction utility for the specific, maddening problem of a yellow or warm cast on an image - the kind you get from scanned photos, old film, bad white balance, or a generator that decided everything should look like 1974. Feed it an image, and it auto-levels the color channels to strip the cast and pull the neutral tones back.

It's essentially a full-featured auto-adjust engine wearing a simple name. Under the hood it splits the image into color channels, applies histogram-based auto-levels to each, optionally rebalances R/G/B via gamma, then layers on brightness/contrast/saturation adjustments - and finally blends the result back with the original by the strength slider, so you control exactly how much of the correction survives.

Inputs that matter

Everything here is required by the schema, but you'll touch maybe four of them:

  • strength - 0–100, default 100. How much of the correction to apply. At 100 it's full auto-levels; dialing to 40-60 is often nicer for a subtle cleanup.
  • mode - the interesting one: RGB, lum + sat, luminance, saturation, or mono. RGB auto-levels each color channel (the default yellow-cast fix). luminance levels only lightness in LAB space (great when you want to keep the color mood but fix flat contrast). mono is a straight grayscale auto-level. Think of mode as "what am I actually correcting."
  • red / green / blue - -100 to 100, per-channel gamma rebalancing. The surgical instruments for a stubborn cast that auto-levels can't fully kill.
  • brightness / contrast / saturation - the conventional enhancers, -100 to 100.

Output is a single image tensor, same dimensions as the input, ready for SaveImage or to continue down the graph. It handles batched image tensors too, so it'll correct a whole batch of frames in one pass.

Installing

Same pack - ComfyUI Manager (search Leon) or:

cd ComfyUI/custom_nodes/
git clone https://github.com/l3ony2k/comfyui-leon-nodes comfyui-leon-nodes
pip install -r requirements.txt

Restart, and it's just there under Leon_Utils. No key, no downloads, no API - it runs on Pillow and numpy locally. This is the one Leon node you can hand to a friend without also explaining billing.

Common issues

Because it's a local Pillow pipeline, there's very little to break. Two habits to avoid: cranking strength to 100 on an image that's only slightly warm will over-correct into a cold/teal shift, so ease it down; and if your input has a meaningful alpha channel, remember the node restores RGBA at the end, so downstream nodes that don't expect alpha may need the usual handling. The single best workflow tip: run it in RGB mode for the yellow cast, then luminance mode if the result still looks flat - two passes beat one aggressive pass every time.

CategoryLeon_Utils

Inputs (9)

NameTypeDefaultDescription
imageIMAGEβ€”
strengthINT1000–100β€”
brightnessINT0-100–100β€”
contrastINT0-100–100β€”
saturationINT0-100–100β€”
redINT0-100–100β€”
greenINT0-100–100β€”
blueINT0-100–100β€”
modeCOMBORGB5 options: RGB, lum + sat, luminance, saturation, mono

Outputs (1)

NameTypeDescription
imageIMAGEβ€”