I trained using natural language for the captions rather than tags so you'll probably get the best results going that route with your prompts. I mostly trained on three specific image types. Images from Danganronpa the Animation, sprites from Danganronpa 1, and swimsuit sprites from Danganronpa S. You can push the image style between anime and game by specifying either 'screencap from the anime Danganronpa: The Animation' or 'danganronpa sprite style'. In theory at least. In practice I'm finding it something of a coin flip.You might get the correct clothing for Sayaka by just mentioning 'uniform from danganronpa 1' or 'swimsuit from Danganronpa S'. However, it's generally best to be specific with something like the following: 'Sayaka Maizono wearing her swimsuit from Danganronpa S. She is drawn in the Danganronpa sprite style. Her swimsuit is a light blue bikini with a white ruffle, and a teal sarong tied around her waist.' or 'Sayaka Maizono wearing her uniform from Danganronpa 1. Sayaka has a blue skirt and her legwear consists of navy thigh-high socks, the edges striped with a white line, and dark brown Mary Janes with silver buckles. Her school crest is printed in gold on her collar flap and her left sleeve.' I'm planning to make more Danganronpa characters over time using a similar process.

Files with SHA256: bc944daf4e8cebe7033682c843e23504c42ef4fe7b242ece0573cd66a4ab5795