Overall

This lora trained by Atlus official setting artbook, training data have been filp in order to do data enhancement. During training the loss value didn't drop well, but the effect seems good. Human ??????? performs not well. Recommanded to combine with Controlnet and Adetailer. If possible I will try other learning rate in order to drop the loss and hopefully could improve the human ??????? drawing performance.

Recommanded weight

0.8-1

Trigger word

PersonaCruncher,

Other

私心说一句，练这个lora其实是看到ave mujica的新消息，灵光一现，都是戴面具，ave mujica又何尝不是一种persona（）p5的总攻击结算界面确实很搭配大祥老师。2024年了我还在go，还在go！