我用别的吟霖Lora在三次元模型上跑出来效果很差，所以我就针对写实模型训练了这个Lora。用这个Lora跑cosplay照片泛华能力还行，调高权重以后该有的特征也大差不差。

I used a realistic model to train this Lora, it can generate high quality Yinlin's cosplay pictures.

推荐权重/Recommended weight: 0.7~1