AniEdit (Flux Kontext)
Details
Download Files
Model description
A comprehensive lora for editing anime images, based on Flux Kontext.
VERY EARLY VERSION.
Works with quants.
Can:
- Declutter/Remove text
- Zoom into specific details
- Create a composite from images of multiple subjects
- Change poses
- Turn chibi/game avatar into a full illustration
- Generate from an empty canvas (from tags or natural language)
- Render from a sketch
- Change specific features (colors, hairstyle)
Trained on 1000~ images.
WORK IN PROGRESS
These features don't really work yet:
- Rough pose sketch to character
- Rendered to sketch
- Background manipulation (changes)
- Pixelate
- Generate using art style references (stitched references)
- Fill-in pasted images/PNGs seamlessly
- Costume Swaps between two characters
You can message me for the dataset link if you want to make something similar (about 700 image edit pairs)