Initial test version, it was trained low quality publicly available images and screenshots of the episodes with the intent of capturing the style and not so much specific details. This seemed to work ok but will need to create a new version once I learn how to better guide the style.
