<h1 id="heading-1099">LoCon</h1>Use these keywords to activate "Kyoto Animation screencap" "Anime screencap" recommended weight: 0.5~0.7Suitable Model: anything, anylora, dpepmkm, detailed project v4HiRes Fix: Latent - AnimeSharp - ESRGAN<h1 id="heading-1100">LoHA</h1>Use this word to activate "kyoto style"Recommended weights : 0.4~0.7(1 works too)Hires Fix Upscaler: R-ESRGAN 0.7 Denoise or 0.4, i prefer 0.7 usually. You can use anime sharp or any upscaler of your own choiceSuitable Model: AnyLoraThe finalized version is better than the experimental and the locon.

Kyoto Animation screencap Artstyle

Trained as a LoHA, different base model. Same dataset, 768x768 and due of colab limits it wasn't fully trained.

Files with SHA256: 04ad78a1945c8a9bf95646974f50ead6220d9898339559bbfe9004e3c0feeb8d