Karma Tone Curves

If you've ever graded a photo in Lightroom, you know the tone panel: lift the shadows, recover the highlights, add a touch of midtone contrast, maybe tint the shadows teal. Karma Tone Curves is that panel as a ComfyUI node - independent shadow/midtone/highlight brightness, an S-curve contrast control, black and white point clipping, and split toning, all on a single IMAGE in, IMAGE out. It's the surgical half of this pack's post-processing chain, for when Kolors is too blunt and Film Emulation is too automatic.

Under the hood it's doing what a real tone curve does, just organized as zone controls instead of a graph you drag. The shadows, midtones, and highlights inputs (each -1 to +1) adjust brightness in those ranges independently - negative shadows crushes blacks, positive lifts them; midtones behaves like gamma; highlights pulls down or pushes up the bright end. midtone_contrast (-1 to +1) applies an S-curve, so positive gets punchier and negative goes flat and airy. black_point (0–0.3) raises the floor for that faded, matte look; white_point (0.7–1.0) drops the ceiling to tame blown highlights. The clever bit is the implementation: it blends the zones with cosine weighting so the transitions stay smooth, which is specifically to stop banding - the classic failure mode of naive per-zone adjustments.

Split toning rounds it out: shadow_tint_hue/shadow_tint_strength and highlight_tint_hue/highlight_tint_strength let you tint the two ends of the range with any hue, exactly like Lightroom's split toning. Here's the one number system to internalize: hue is normalized 0–1, where 0 is red, ~0.33 green, ~0.6 blue, ~0.83 magenta. That means the famous teal-and-orange grade is shadow_tint_hue 0.52 (teal lives right before blue) with highlight_tint_hue 0.08 (orange, just off red), both strengths around 0.1–0.15.

Installing it

Search "KarmaNodes" in ComfyUI Manager, or:

cd ComfyUI/custom_nodes
git clone https://github.com/KarmaSwint/ComfyUI-KarmaNodes
cd ComfyUI-KarmaNodes
pip install -r requirements.txt

Restart ComfyUI. No model downloads - it's pure NumPy pixel math.

Recipes worth stealing

The README has a handful that genuinely work. Faded film: black_point 0.05–0.1, shadows +0.1, midtone_contrast -0.1. Punchy modern: midtone contrast +0.3–0.5 with shadows slightly crushed. Matte finish: black point 0.05–0.08 and white point 0.92–0.95. High key: lift everything - shadows +0.2, midtones +0.1, highlights +0.1.

The trap to watch is subtlety on the wrong axis. Because the zones blend smoothly, big single-zone moves are safe from banding but can still look plastic - the reads of "muted," "matte," and "washed out" all live within about ±0.1 of zero, so nudge, don't slam. And remember its place in the pipeline: the pack suggests tone curves first, then color, then lens FX, then grain. Getting the tonal structure right before you touch color and grain is the whole difference between a grade and a filter.