Schmuser Realism: Chromatic Aberration Slider - Krea 2
Adds real lens fringing to your images.
What it does:
Real camera lenses don't focus every color on exactly the same spot. Toward the corners, the red and blue parts of an edge drift apart. Bright edges against dark areas get a red/cyan double contour, and backlit branches or hair get a purple glow. Krea 2 renders an optically perfect lens, so its photos can look a bit too clean, too "rendered".
This LoRA brings that lens character back:
The effect grows toward the edges, like a real lens: the center (usually your subject's face) stays sharp and clean.
Purple fringing on bright edges: window frames, backlit branches, neon tubes, white text on dark.
A very slight corner softness that goes with it.
Use cases
Analog / film / lo-fi realism: disposable camera, vintage glass, 90s snapshots.
Neon and night photography: cyberpunk streets and signs, where fringing makes light feel physical.
Street, architecture and fashion photos shot "on a wide-angle lens".
Glitch / RGB-split looks at high strength.
Making AI photos look less AI, adding the small optical flaws a real camera always has.
Why not just prompt "chromatic aberration"?
Prompting changes the whole photo. The lens words act like a new prompt: a different street, dress and sign. The fringing you get is random, from a little (street) to none (night).
The slider is controllable. 0.5 is a hint, 1.0 is clearly visible, 2.0 is a cheap plastic lens. No prompt tweaking, no rerolling.
It's physically placed. Fringing grows toward the borders and sits on high-contrast edges, just like a real lens, not as a uniform filter.
It keeps your image. Used with the workflow below, the same seed gives largely the same picture (face, pose, layout), just through a different lens.
How to use
Recommended: Krea 2 Turbo, 8 steps, cfg 1, euler / simple, with two samplers:
1. Sampler 1: steps 0 to 3 on the plain model. This lays out the composition.
2. Sampler 2: steps 3 to 8 with the slider. This adds the lens.
Set the amount with strength_model on the Load LoRA (model only) node:
0.5 - subtle, "good lens with a bit of character"
1.0 - clearly visible on edges and near borders
1.5 - strong vintage look
2-2.5 - heavy rgb split / lofi look
Change point (
start_at_step/end_at_step= 3): 2 gives a slightly stronger effect but small details can change. 4 is the most faithful but weaker.Use positive strengths only. Negative values don't remove aberration (Krea 2 has none to remove). They just change the image.
Works with other LoRAs. Put the slider last in the chain on the sampler-2 model.
Best subjects: high-contrast edges away from the center, such as backlit hair and branches, window frames, railings, neon, text, chrome and wet highlights.
Simple use (one sampler): you can also load it like any LoRA. The fringing works the same, but at strength 1+ the slider also nudges the composition (pose, background details). If that's fine for you, start around 0.5 to 0.8.
Strengths and weaknesses
Strengths
Real, physically-placed chromatic aberration that Krea 2 can't produce by prompting.
Smooth control from subtle to heavy with a single number.
With the two-sampler workflow, the image keeps its composition: same face, pose and layout.
Leaves the center of the frame clean, so portraits stay sharp where it matters.
Works on photos, night scenes and neon. On flat illustrations it only adds a faint fringe to outlines.
Weaknesses
Loaded the normal way at 1+, it shifts composition.
It can't clean up aberration, and negative values just alter the image.
A soft, evenly lit scene or a center subject on a plain background shows little effect.
depending on the seed: background objects, small text and signs can change (they are re-drawn in the last steps).
Slight corner softness and warmth at 1.5 to 2 come along with the effect.
Made and tested for Krea 2 Turbo (also loads on Krea 2 Raw, untested there).
Thanks for checking out my LoRA and I hope you find it helpful in your artworks.
Feel free to leave some feedback or let me know what could be improved. 🫶