Images used to train were 100% women, 95% nsfw, and 95% empty or simple geometric shape backgrounds. Will have trouble generating men, and will need to prompt clothes and backgrounds.
