First version, training images have been made with SD 1.5 (a personal merged model). Lora was trained with 2 datasets:
1) style data set
2) character data set
A virtual girl named Michaela, trained at 800 steps, ZImage base
