If speaker embedding is not added to the encoder input, will it affect the model effect?

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

Open qq547276542 opened 4 years ago

qq547276542 commented 4 years ago

If speaker embedding is not added to the encoder input, will it affect the model effect? Do you have any relevant experiments? Thx！

auspicious3000 commented 4 years ago

No fundamental difference