缠枝/QBU-88

This LoRA is only for Anima 2.9B.

If you want to use it in anima-base-v1.0, remap the LoRA layer keys back to the 1.0 indices

Character LoRA for Chanzhi / QBU-88 from Girls' Frontline

Trigger words: chanzhi, qbu88

Recommended settings: model 1.0 / clip 1.0; euler + simple scheduler, steps 35–40, cfg 5.

Training info: trained with Anima-Standalone-Trainer, standard LoRA, on Anima 2.9B native 40-layer architecture. Dataset: 99 images (10 primary ×4 repeats + 59 secondary ×1). 15 epochs, 1485 steps.