Clone a voice in 5 seconds to generate arbitrary speech in real time

github.com

50 pointsroseway43 comments

How does output compare to Mozilla TTS or Google's new customizable TTS: https://sample-efficient-adaptive-tts.github.io/demo/

The latter requires a few minutes of audio to tune it to clone a voice.

I'd say the 10 second trained clips towards the bottom sound much better than the clips from the YouTube video of this project.

Ironically demo YT clip is "relative_loudness": "-17.739dB", aka cant hear it. Not great for someone doing sound stuff.