https://github.com/scraed/LanPaint
- Inpainting mode: Image First
- Consistency mode: Prompt First (CFG=1 does not work)
- Supports fine-tuned models
- Supports face swap tasks
More use cases: https://github.com/scraed/LanPaint/issues/12
LanPaint uses Langevin Dynamics as "thinking" steps, which digs deeper into the diffusion process and allows the model to generate more consistent results.
Works instantly with almost any model (SD 1.5, XL, 3.5, Flux, HiDream, or custom LoRAs) and ControlNet.