changed base model from epsilonPred075 to illustriousXL_v01
Lora for Visi, using at weight 1. For some reason a lot of checkpoints make her more "anime"-like (maybe just my prompts), but the recommended one works pretty well.
