Updated: Oct 6, 2024
style
Once again, I feel like I have no idea what I'm actually doing. I decided to try CivitAI's new training system and this is the results. I used 9 artworks from the game Slay The Spire without captions and CivitAI's default training settings, using an anime SD1.5 checkpoint as a base (ask them which one it is, idk, prolly anylora).
I've gotten great feedback from my peers on this, and some helpful advice for if I decide to try for a v2. The lack of captioning is hampering it, and just 9 images isn't the greatest base to work with, so it's pretty rough, but it works. There are no trigger words... captions might have been useful here too. I recommend rendering with Cartoon and Anime style checkpoints that don't use heavy outlining.