The purpose of this model is to create "normal" as also horror clowns,
Trained on 35 images of (horror) clowns.
Best between 0.5 and 1.