On: The uncanny failures of A.I.-generated hands

[spaced-out]

It is possible to generate good quality hands with Stable Diffusion with ControlNet. Here's a video talking about generating different hand positions from models created in Blender. There's also discussions about this on r/StableDiffusion.

https://www.youtube.com/watch?v=ptEZQrKgHAg

Now from personal experience, it's not as easy as it seems in the video. It still takes a lot of tweaking and many iterations don't come out well, but hands are definitely not an insurmountable problem.

[prox → spaced-out]

Depth controlnets have been the most helpful sofar, they give space directions the best and untangles the good from the bad imo,

[loufe → spaced-out]

I mentioned it in another response as well, but here's the original implementation repo for anyone interested:

https://github.com/lllyasviel/ControlNet