This LoRA is designed to reproduce the character accurately and consistently using a single trigger word,

while minimizing interference with clothing, background, pose, and overall style.

It maintains character identity across a wide range of facial expressions(neutral, serious, glaring, smiling, angry, embarrassed, surprised, etc.).

this LoRA performs best at weights around 0.78–0.81,with 0.80 being the recommended default.