🌈 CYH Post Process | Chromatic Aberration

Chromatic aberration is the colored fringing you see at the edges of a photo - red bleeding one way, cyan the other - the thing cheap lenses do and expensive ones are designed to eliminate. Which is exactly why it reads as "photographic" when you add a hint of it back. This node adds it with per-channel distortion and a few presets so you don't have to think too hard.

It's one of the four post-process nodes in the Chye ComfyUI Toolset, wired in the usual spot: KSampler → VAE Decode → CYH Post Process | Chromatic Aberration → Save.

How it works

Real chromatic aberration isn't a color shift - it's each color channel being magnified slightly differently by the lens. This node does the same: it splits the image into R, G, B and applies a barrel distortion to each channel independently using a classic lens model:

r_distorted = r * (1 + k1*r^2 + k2*r^4 + k3*r^6)

Each channel has its own k1/k2/k3 coefficients (the k1_r, k2_r... inputs), which is how you get red edges fringing one way and blue the other. The remap is done with OpenCV, and the distortion maps are cached so re-running with the same parameters doesn't rebuild them.

The inputs that matter

Three inputs to actually touch, plus an IMAGE in:

• preset - none, vintage, modern, extreme, custom. Vintage is the default and the workhorse: red +0.15, blue −0.15, which is the classic subtle lens look. Extreme goes to ±0.25 with higher-order terms - that's the "glitchy" look, more art than lens.
• intensity (0–2, default 1) - a master multiplier on all the coefficients. Turn it down for a barely-there hint; at 2.0 even vintage gets loud.
• quality - fast (bilinear) or high (bicubic) interpolation for the remap. You won't see the difference on small images; high costs a bit more CPU.

The advanced k1/k2/k3 per-channel inputs are only relevant if you pick custom - with any other preset the coefficients are overwritten. center_x/center_y shift the distortion center (default 0.5/0.5), which is the classic trick for making fringing look off-center and organic rather than perfectly radial. Output is IMAGE.

Where it bites

The pack-wide limitation applies again: only the first image in a batch is processed (image_np[0]). And it's a CPU pass via OpenCV - fine for one image, slow for batches.

The other thing to keep in mind is taste: 1.0 on the vintage preset is already clearly visible on high-contrast edges. This is a "small dose" effect. If you're adding it to a photorealistic image and people can name the effect, you've used too much.

Install

ComfyUI Manager (search "Chye ComfyUI Toolset"), or:

cd ComfyUI/custom_nodes
git clone https://github.com/chyer/Chye-ComfyUI-Toolset
cd Chye-ComfyUI-Toolset
pip install -r requirements.txt

Restart after. Needs opencv-python (in the pack's requirements; you likely already have it). Zip installs need .git/.cnr-id (Chye-ComfyUI-Toolset) per the README.

Verdict

Presets plus an intensity knob is the right design for an effect like this - most people want "a little vintage," not a coefficient sandbox. It's a solid, no-frills tool for a specific look. Just remember the batch caveat.