v4 Should have the extra detail of v3, but be less baked than v2, recommended weight is 1, but you might have to lower it for some attention in some checkpoints.V3 is just a color and sharpness improvement + flexibility in expressionsV2 now with a style which resembles Tsuji's modern art a bit more, has trouble with really big badonkas, but has no issues using <a target="_blank" rel="ugc" href="https://civarchive.com/models/44785/venus-body-or-concept">mnemosynekotorin's</a> venus body lora.All generated images in the post for v2 were run with Adetailer after the Hires fix, take it into account! A model based on Tsuji Santa's style, mostly trained on Sonico, Pochacho and some fanart for Muramasa, and CGs from Axanael, if you like this, please support <a target="_blank" rel="ugc" href="https://fantia.jp/fanclubs/86661">Tsuji Santa</a> Model doesn't generate headphones as much but if you prompt "super sonico" she will still have hers even if you put headphones in negatives, its better if you prompt her manually if you don't want her with headphones, strength goes from 0.5ish all the way up to 1.2.If you like what I do please consider supporting me via <a target="_blank" rel="ugc" href="https://ko-fi.com/bluvoll">Ko-fi</a><h2 id="heading-502"></h2>

Tsuji Santa | Artist style

Same dataset, cleaned tags, more flexible expressions

Files with SHA256: 1fb0dd8444c8a46cae7814a3f9a8e265e2c75c15cedf246785fc6199fc839d68