Its only trained on other images that look like the ones seen in the cover so I doubt its useful at generating at anything other then the character.