I noticed that the model uses Hifigan as the vocoder. How was the hifigan model trained? On what data?
· Sign up or log in to comment
