Christian Schäfer

Results 72 comments of Christian Schäfer

Hey, glad you are interested. GAN worked but the vocoded wavs were pretty noisy, although I probably did not train for long enough :-) . Currently I am doing some...

Quick update on GANS, with least squares GAN (https://arxiv.org/abs/1611.04076) and conv discriminators I got about equal quality as with a simple L2 distance objective, albeit no improvement over it.