Version 2 This LoRA was trained on 21 handpicked images from gelbooru, zerochan, and pixiv with Anything v4.5. In this version, I reduced the number of fanart and added more screenshots from the anime into the dataset. This LoRA might be a little overcooked, but I think that its way better than the previous version.
Version 2
This LoRA was trained on 21 handpicked images from gelbooru, zerochan, and pixiv with Anything v4.5. In this version, I reduced the number of fanart and added more screenshots from the anime into the dataset. This LoRA might be a little overcooked, but I think that its way better than the previous version.