/vt/ - Virtual Youtubers

Also a curious thought. Those who have their AI generated models from an imperfect angle, how difficult would it be to generate an A pose image using a model built on their character and something like img2img of a more generic A pose you usually see in vtuber art? Seems like people here already made custom trained models of the talents, have you tried combining them in img2img? Maybe you'd need to do some lite painting on the A pose image to help the AI match the elements though.