This version was trained on collages of characters with (mostly) realistic backgrounds. The images were made by members of Pinterest using "cutouts" of both existing and original characters, but also with a technique utilizing pieces of anatomy allowing figures to be posed.