This model has four tokens in memory each one associated with 30 to 40 different style images, there is a huge variety of outcomes and no overtraining means it will respond to any input perfectly.
Each token starts with wjq for unicity purposes:
WJQWIZARD, WJQKNIGHT, WJQDEMON those are self explanatory, even if maybe some classes might be associated with affine concepts; like a wjqwizard could give you the image of a witch.
WJQJESTER is based on a creative mix of colorful body art, creepy puppets, carnival masks, clowns.
WJQTITAN will likely give you a mountain setting with a huge entity in the middle, there are several possible types.
WJQAUTOMA for ancient robots, scarecrows, animated statues (new in v2).