Tried to make a lora with some ai generated anime images . This is just for testing . It seems to work ok with <a target="_blank" rel="ugc" href="https://civarchive.com/models/119032">unaestheticXL</a> embedding. The images needs to be created with 1024x1024 . 512x512 will give bad results. This lora is over trained (43272 steps/28 batch size= 1545 steps) , so around 0.7 to 0.8 weight seems to give better results.Prompts for the showcase images are copy pasted from civitai images.

DMZXL

<ul><li>Was done without refining the dataset. </li></ul>

Files with SHA256: 58c8bf16491f0005d5395af1152da438e490a2f968b43c59348475d0ad1562f8