search

auspicious3000 / autovc

AutoVC: Zero-Shot Voice Style Transfer with Only Autoencoder Loss
https://arxiv.org/abs/1905.05879
MIT License
976 stars 207 forks source link

Where is the "Deconv Layer" in figure 3(d) in your paper? #83

Closed gnipping closed 3 years ago

gnipping commented 3 years ago

I retrain the model in VCTK corpus following the training steps in your paper. However, I only get noise in my generated voice. I have a quesion in the file "metadata.pkl" in your project. Does the mel feature matrix in this file has already passed "Deconv Layer"?

auspicious3000 commented 3 years ago

No.

gnipping commented 3 years ago

So is the "Deconve layer" in your paper already include in "wavenet_vocoder"?

auspicious3000 commented 3 years ago

Yes

gnipping commented 3 years ago

OK, thank you very much!