Sometimes the reference voice would occured at the systhesized voice

#8
by weiren119 - opened

Your works are really awesome.
I try to convert the chinese to fr. It get a good performance.
I connect it with whisper&pyannote/ some translate method to convert multiple speaker to desired language.

But sometimes the reference voice would occured at the systhesized voice.
In your experience, how should we prevent the problem.
Thanks a lot.

Reference voice should be clean (denoised but relative to speech), with no big silences , also not too short for best output result.
Edit : with last 0.19.1 reference voice should not be heard at synthesised voice .

gorkemgoknar changed discussion status to closed

Sign up or log in to comment