/vt/ - Virtual Youtubers

>>75309519

>>75309800
I tried prompting without adetailer, the face tone seems to still mismatch with the body. Could be the model's training on how it handles low-light gens. Funnily enough, dim lighting was easily achieved on kohakuXLepsilon without the use of loras so I do think this lighting stems from animagine3.1