gessyoo

Thanks, I'll give it a try with different locations.

Someone pointed me to this post: https://www.deepfakes.club/forums/topic/test-series-comparing-128x128-trainers/. The writer compared the current 128x128 models using the same data sets. He recommended the DF model from DeepFakeLabs and the OriginalHighRes. The...

I defer to @andenixa since I'm a python novice, but I've made a dozen or so deepfakes testing the various models. I think the optimum batch is the largest batch...

andenixa, I'm aiming for the best possible results regardless of the training time required. You're suggesting, as discussed above, that the model could be adjusted with IMAGE_SHAPE = 256, 256,...

Windows 10 reserves about 2 GB of video memory, so even 8 Gb. cards only have about 6.5 Gb. available to CUDA. Ubuntu doesn't have this problem. There was a...

Thanks for responding. I did try to build it with CMake GUI and Visual Studio a while back, but the build failed. I will make another attempt to compile and...

I'm getting this error in a local install with a 30x GPU. I've tried the potential fixes above, re-created the environment, etc. The 1.5 models load and run as they...

> https://www.microsoft.com/en-us/research/project/vasa-1/ I've made some videos that look better than those VASA-1 examples, but I'm also looking forward to the release of the pre-trained audio model, especially since M$ won't...

> > > https://www.microsoft.com/en-us/research/project/vasa-1/ > > > > > > I've made some videos that look better than those VASA-1 examples, but I'm also looking forward to the release of...

> > > > > https://www.microsoft.com/en-us/research/project/vasa-1/ > > > > > > > > > > > > > > > > > > > > > > > >...