@ayakashiworks 妹るーてぃん! Style/Chars

Initial. (experimental)This LoRA uses the same training images and captions as the Anima version.<pre><code>year 2022: 29 images
year 2026: 64 images</code></pre>When using this LoRA with MOSS Image SFT, a LoRA weight of 2.0 or higher is recommended.However, 3D-style rendering may still be unreliable even with this LoRA. If you specifically want 3D-style results, use the base model instead (using weight 1.0 or higher).LoKr, MOSS Image Base, dim32/alpha32, CAME, UNet LR 3e-05 cosine 1,000steps, 93 images (512² res).Captioning: Booru tags only.

Files with SHA256: f40481230e736a46ee860c21d1205c5ff461636bdd7a5a5c7101c5f054f2aa5a