search

pyannote / pyannote-audio

Neural building blocks for speaker diarization: speech activity detection, speaker change detection, overlapped speech detection, speaker embedding
http://pyannote.github.io
MIT License
6.38k stars 784 forks source link

Speech Separation cracking the volume too high #1770

Closed ajtopper closed 1 month ago

ajtopper commented 1 month ago

Tested versions

-Reproducible in pyannote 3.3.1 - speech separation ami 1.0

System information

ubuntu 24 - pyannote 3.3.1 - speech separation ami 1.0

Issue description

Hello folks,

I'm using Speech Separation ami 1.0 and when there is separation by speakers, the final files end up with very high volume, with the entire waveform being distorted and clipped.

Is there any way to control this output volume?

Here have a example: https://drive.google.com/file/d/11Ber2v9HTKUaHWs4hCEbgoo9s4gPBlKC/view?usp=sharing

Thanks.

Minimal reproduction example (MRE)

https://colab.research.google.com/drive/1Yr2xGz-UL4gpElyLGV8HzE5bfhGgOFIN?usp=sharing

clement-pages commented 1 month ago

Hey @ajtopper, your issue should be fixed by #1730.

ajtopper commented 1 month ago

Hey @ajtopper, your issue should be fixed by #1730.

Works perfectly, thanks!