宇佐见莲子的LoRA模型,属于是我尝试训练LoRA模型的初作,但之后训练的LoRA的质量似乎比这个还不如呢(笑)。
The LoRA model of Usami Renko. This is the first outcome of my attempt to train a LoRA, although the following works are not as good as this one.
一般权重在0.8-1。我强烈建议加入一些外貌的详细prompt来稳定该模型的表现,以下是我常用的prompt:
The recommended weight is 0.8~1. I strongly suggest adding more detailed prompts on appearance to stabilize the outcome. These are the prompts I often use:
renko, black victorian hat, short black hair, black capelet, red tie, white shirt, black pleated skirt, white socks, black dress shoes,