this model reproduces a very consistent character as long as you use DPM++_2M_SDE_ with KARRAS in a 3.5-5.5 CFG and 32-40 steps..... the lora was trained on the Lustify v8 Apex checkpoint but does extremely well generating images on the Lustify v7 checkpoint. (((this character data set was created purely from Ai derived renders, and spent the time creating multiple headshots using VeniceAi to regenerate the face in different angles.)))