anhnh2002 / XTTSv2-Finetuning-for-New-Languages

60 stars 17 forks source link

Dataset size #18

Closed lukaszliniewicz closed 2 weeks ago

lukaszliniewicz commented 2 weeks ago

How much audio have you used to obtain decent results?

anhnh2002 commented 2 weeks ago

How much audio have you used to obtain decent results?

I've found that approximately 100 hours of audio data are necessary to achieve satisfactory results.

desis123 commented 1 week ago

I've found that approximately 100 hours of audio data are necessary to achieve satisfactory results.

That 100 hours need to be one person's voice ? Or I can use multiple person voices to make that 100 hours ?