First attempt at a style. Based on one of my all time favorite games: Slay the Spire. This model was trained with NO CAPTIONS. Meaning is has no idea what Ironclad, The Silent, Watcher, or The Defect are. I may retrain with hand typed captions so the specific characters are understood but for the first pass I mainly just wanted the "style" of the art rather than the characters themselves. The dataset was cropped images of all the card art from the game files. (excluding all powers as they were round shape and cropped very poorly). Trained 15,000 steps