Open MotherSoraka opened 1 year ago
Unfortunately, right now they can't. The models expect the images to be in a different range than we currently support, and we have no way of automatically detecting model architecture in NCNN right now. We have discussed some workarounds though, so maybe something will be implemented in the future. Or, maybe I can just add a pytorch version of it.
Hi, I absolutely love this piece of app! One of the best things ever happened to humanity!
I also love Real-CUGAN, but i can't get it to work in ChaiNNer. Only one of the models "up2x-no-denoise" works, rest of the models output black images.
Any ideas how i can fix this?
Thank you!