Trained on the art of SketchGoblinhttps://www.instagram.com/thesketchgoblin/?hl=enhttps://x.com/SketchgoblinHis art focuses on D&D characters in a flat style against a white background. In captions, I've included races and classes, but this LORA is not enough for the model to know what "warlocks" or "shadar-kai" are. However, adding "warlock's clothes" instead, for example, helps. Still, try to describe things in a way the model knows.Also there is a lot of weapons in the dataset, but pony is weird with weapons, so it's almost like it's better to add it to the negatives, but experiment with that.Every caption also has a "dnd" tag in it, but using it is not necessary, though it alters the result in some way.
