Output is a bit shaky, how to fix that?

[Souvic]

Thanks for the great work and making code with all weights available! Really appreciate it..

Can you please guide me on how to improve the output further? If we change the vocoder to HiFIGAN V2 or train on more data, how do you think output will change?

Also, how much time does it take to train on train-100 data from librispeech?

[RF5]

Thanks for your interest in our work -- there could be many ways to improve things, I'm not sure on what the best way is to make it better. For some of your suggestions, here's how I predict they might change the output:

• If you use HiFiGAN V2, I think it will do worse than HiFiGAN V1. V2 has much fewer parameters and scores much worse in the original HiFiGAN paper, so I do no think using V2 will improve it.
• Like with just about every model, if you train on more data it will probably improve things.

The training time for training on train-100 data from librispeech took around a week or two on 3x Quadro RTX 6000 GPUs. I hope that helps!