Speech Not Clear

[MMingabc]

I am using my own dataset which is a Chinese professional tts speech dataset of 8k utterances to train bert-vits2. The generated speech is not clear. It sounds like the speaker but the content cannot be understood. It's just mumbling. What could be the problem? Thank you. And I am using 16k sampling rate.

[AngelGuevara7]

Same problem here for spanish! The alignment matrix looks good but the audio it's just mumbling. I'm using 44,1k sampling rate.
Did you solve it?

[starmoon-1134]

This maype help you.
#245 (comment)

[AngelGuevara7]

Thanks! I will take a look at it.