/vt/ - Virtual Youtubers

there's still a lot of things happening here that i don't fully understand.
i trained the lumi LoRA on a different model but i thought i overtrained it. 33 epochs barely did anything to the hair, 66 epochs started to do the hair better but started introducing random noise and artifacts in the generated pic.
but then i change to a different model and the LoRA also works here, in fact, it seems to work even better with less noise and artifacts??

it really is like magic and the more i do the less i understand.

anyway have some kid lumi