I have VERY limited knowledge on stable diffusion so take it with a grain of salt. I figured that uncensoring actual pre-rendered animations would be nearly impossible, but very viable with games using live2d for example. I just don't know how to prompt it to uncensor the baked texture and i can't find a guide anywhere... Imo it's worth a thread
In my experience, you can achieve fair results using A111 / Vladmandic, or comfyui if you want absolute control.
You just need the right model, and maybe even some LORAs to add a certain style for sake of consistency.
Best practice is masked inpainting with img2img, and just prompting what's needed.
You could also try deepcreampy, it's standalone, but it's pretty outdated and results will vary.
IMO, for AI to be worth your time in the current moment of advancements, you need a high end PC and a ton of time to waste.
My last venture into AI demosaic can be
seen here. You can clearly see the problem with consistency, and the problems I ran into later on.
Being a graphics artist is still the hands down best way to demosaic textures.