What are the differences between the model hosted in this repo and the true model?

auspicious3000 / autovc

AutoVC: Zero-Shot Voice Style Transfer with Only Autoencoder Loss

https://arxiv.org/abs/1905.05879

MIT License

983 stars 207 forks source link

What are the differences between the model hosted in this repo and the true model? #56

Closed ngulya closed 3 years ago

ngulya commented 3 years ago

In other issues I have seen that a request to IBM (through you?) is required to get the full code. Where to write if it's so?

auspicious3000 commented 3 years ago

They are the same.

ngulya commented 3 years ago

Okay, a question from the other side) I noticed that some inaccuracies were added to this code (not saving the model, adding noise, another metadata format), the question is why?

auspicious3000 commented 3 years ago

The code is made as simple as possible so that it does not cause unnecessary confusion. It is my personal habit to add a little bit of random noise to training audio. Free feel to remove it. The test and train metadata have different format and it is very easy to figure out.

ngulya commented 3 years ago

Thanks for the answer and for the paper!