Hifigan model #3

by tshmak - opened

I noticed that the model uses Hifigan as the vocoder. How was the hifigan model trained? On what data?