Oh, no I'm happy to explain if you're interested.
I have a personal ComfyUI setup on my pc (a free frontend for generative ai for images). I set up a workflow to take the censored image and weighed it heavily to keep the shape and color as close as possible with just enough leeway to actual do the uncensoring part. I ran about 16 runs of each one to get a good one that closely matched the original, then went to town in photoshop to get it properly resized and replacing them, for most. There was a bit of recoloring and smudging work needed, specifically for the female parts, to get it to the right skin tone.
All in all, though, I don't necessarily think it can be fully automated. You can get about 90% of the way there using AI as a tool, but you'll still need to isolate the area, write up the correct keywords for each, cherrypick the right image, then resize it and fit it back into the original image in the editor of your choice. Like any tool or editor, it's just a manner of learning how to use them, then the process gets cut down timewise greatly. That and a good enough pc to run an art model. You can probably get by on 4gb of vram for an art one, but I'd recommend at least 8 to not make it slog while the genning process.
