KevinMIN95 / StyleSpeech

Official implementation of Meta-StyleSpeech and StyleSpeech
MIT License
241 stars 39 forks source link

How to improve the synthesized results? #19

Open sanjeevani279 opened 2 years ago

sanjeevani279 commented 2 years ago

I have trained the model for 200k steps, and still, the synthesised results are extremely bad. The sampling rate I have used is 22050 Hz and the batch size used is 16.

loss_curve

This is how my loss curve looks after 200k steps. Can you help me with what can I do now to improve my synthesized audio results?

chazo1994 commented 2 years ago

@sanjeevani279 I have same problem with 22050 hz, while 1600hz is ok. Did you resolve this problem ?

Summerxu86 commented 2 years ago

@chazo1994 Are you using 1600Hz and the batchsize is 16? How is the synthesis effect?

chazo1994 commented 2 years ago

@chazo1994 Are you using 1600Hz and the batchsize is 16? How is the synthesis effect?

@Summerxu86 I train two model 22050khz and 16khz, both use batchsize 48. Model 16k is faster convergence, and the synthesized audio at 200k step of model 16khz is much better than model 22k.

Aliraheem commented 3 months ago

@chazo1994 Are you using 1600Hz and the batchsize is 16? How is the synthesis effect?

@Summerxu86 I train two model 22050khz and 16khz, both use batchsize 48. Model 16k is faster convergence, and the synthesized audio at 200k step of model 16khz is much better than model 22k.

What vocoder did you use? in the case of 16k and 22050? Did you use a different pre trained vocoder for each sampling rate?