The LoRA was trained on 100 images, consisting of 61 manga pages and 39 illustrations.
The dataset consists of:
year 2026: 13 imagesyear 2025: 26 imagesyear 2024: 45 imagesyear 2023: 11 imagesyear 2022: 5 images
Known Issues
blue skin: Due to incorrect tagging of some training images, the LoRA does not reliably respond toblue skinin v1.0.blush,sweat: These elements may appear even when not specified. Add them to the negative prompt when they are not desired.
Description
Initial.
Anima base-v1.0, LORA, dim32/alpha32, CAME, LR 3e-05, cosine 1000 steps, 100 images (1024²)

