ClariNet

ClariNet copied to clipboard

there is a lot of noise in the high frequency?? do u have any solution?

In my ClariNet repository, I found that using only generated means reduces noise. https://github.com/dhgrs/chainer-ClariNet/commit/576060561ba8b5a7b5e03d5c01aed4213cfbb6df

It means that predicting values instead of probability distributions.

@dhgrs although it sounds unbelieveable, i will try it

I think one possible reason is that there's no windowing process in STFT. https://github.com/ksw0306/ClariNet/blob/b03f99a64087e6eaf7682536b04379f1fe71b38a/modules.py#L117

Without windowing, we will see unexpected high-frequency values in the spectrum domain due to the discontinuity between edges in time domain.

I also test the stft function in the pytorch and a good spectorgram loss is very important. this is the predicted wav: if we listen carefully, there will be some noise in the background

in waveglow and flowavenet, i also found some noise like https://github.com/ksw0306/FloWaveNet/issues/1#issue-378235981 but much smaller than this picture

There is 3 high frequency noise in my synthesis with teacher model. Does anybody meet such kind of issue? Wave file is attached. Thank you. The generate_428934_0.wav is synthesized wav, and the generate_428934_0_truth.wav is the recorded wav used in training. wav.zip

Hi - anyone found the solution for high frequency noise?