Earlier this year I posted a support request for my investigation into StableDiffusion (see https://esotericsoftware.com/forum/d/25752-how-to-export-psd-layers-of-every-pose-by-script/7)
But it ends up to be a more challenging task than expected, and I have to code a lot.
This is what I achieved https://www.bilibili.com/video/BV1fhWzetE4D/?vd_source=3ccdfd13f35470ceba3ca3ca6f1d35d1
Well, it's suboptimal. But however i try, it won't improve any further. So basically that's it. We have to wait for stronger AI.