NVIDIA/waveglow

Is there a way to increase speaking speed via WaveGlow Vocoder?

Closed this issue · 3 comments

Hi,
Is it possible to change speaking rate of synthesis by tuning any parameter of Waveglow model?
Thanks.

you can change sampling rate and train waveglow for lower sampling rate

Try handing this on the TTS side (preferably with Glow-TTS or Flowtron). Increasing the Waveglow sample rate will indeed make changes to speaking speed but they will be rather unnatural, so it isn't highly recommended.

This should be handled on the text-to-spectrogram side.