I do a lot of 3D artwork and I would love to see a model similar to the recovery model for CG. The main difference being that in addition to being able to upscale your render it could also be used at 1x output just to add detail to a 3D render.
The idea came from this blog post and I could see it being incredibly useful for some artists: (bertrand-benoit (dot) com/blog/ai-finishing/)
Especially those trying to increase their render speeds at the cost of quality. The examples in the article show how Magnific can be difficult to tame in terms of directing it, which is obviously a challenge. I feel like you all have done a good job with the Recovery model in that regard as it’s very good at upscaling low res images and keeping the source mostly intact without a lot of weird deviations.
