ParallelWaveGAN icon indicating copy to clipboard operation
ParallelWaveGAN copied to clipboard

Published 20 hours ago • verified publisher icon kan-bayashi

Generated Samples Have an Audible Pop at End

Open shawnbzhang opened this issue 4 years ago • 1 comments

I'm curious if anyone else has experienced this problem. When training the MB-MelGAN, I use the batch_max_steps: 8192 as the length of the data. After training, although the audio quality is quite good, I find that there is a "pop" in audio at the end of the 8192-length audio. Here are plots below:

1 2 3

Does anyone have ideas in how to stabilize/avoid this problem? Thanks!

shawnbzhang avatar Feb 27 '21 19:02 shawnbzhang

@kan-bayashi Sorry to bother you, but would you have any thoughts on this?

shawnbzhang avatar Mar 06 '21 01:03 shawnbzhang