It works
Trained on V-PRED. More images that is trained on and not overtrained too much.
sagawastyleVPREDUPDATESDXLvpred10-step00002800.safetensors
