· Use 15 model capture pictures and 15 paintings (512*512) of Anya Forger from Spy x Family to train it by Kohya_ss GUI on 4080 laptop. My first LoRA, please tell me if you find something wrong or how to improve, I will upadte when free~· I found it work better on checkpoint expmix_line_v20 ~ Anything_V3 &gt;&gt; AOM3/Counterfeit_V2.5 · always open mouth and need more expression to train later(Anya's feature!)...· using open pose and depth to change pose seems work· weight ~0.7 · vae kl-f8-anime2 · 用了阿尼亚的512*512的15张模型截图和15张画作在4080手提上用Kohya_ss GUI训练出来的，我的第一个LoRA，如果发现有哪里不对劲或者那里可以提升麻烦告诉我啦，我会抽空尝试更新的~· 个人使用的大模型里expmix_line_v20 ~ Anything_V3 &gt;&gt; AOM3/Counterfeit_V2.5· 老是张嘴，加点表情训练应该会更好毕竟是阿尼亚啊~· 貌似可以用姿势和深度进一步控制下· 权重用0.7上下· vae kl-f8-anime2

Files with SHA256: 254fb4d01fbd1315ab29a9ae1a738a1bb9e5391cc6dc006c27584ca5c67ca7ab