WaveRNN

WaveRNN copied to clipboard

changing downsample factor?

2

What do I need to change in order to increase the downsample factor from 64 to 128? It seems it is not as simple as adding an additional layer to...

rhoposit

I am training from scratch and I have 360 speaker , its been 490k steps and I have heard anything legible yet, its just noise

TheSeeker218

Hi, thanks for the nice work. Preprocess is already finished fo vqvae model (multispeaker) I have a dataset size with 498005 files. If I try to use "wavernn.py -m vqvae"...

raanubis

Hi, Does anyone tried to use d-vector as the speaker embedding instead of the one-hot vector? When I was trying to use d-vector, the utilization of latent embedding vectors is...

shaojinding

Hello, tryed to train from scratch on multyspeaker VCTK corpus, after 800 epochs I see very poorly results in model_outputs wavs, only first second is going ok, but the end...

crankyz

Hello, tryed to train from scratch on multyspeaker VCTK corpus, after 800 epochs I see very poorly results in model_outputs wavs, only first second is going ok, but the end...

crankyz

pretrained model and single speaker adaption

1

hi, can you share a pretrained model and how to transfer learn to adapt to a single speaker?

ak9250

EMA trains faster

12

Hi, Cool project! When I was trying VQVAE I found that using a moving average like described in the appendix trained a lot faster and gave better results! There is...

pfriesch

Unconditional weights / training tips?

2

Has anyone had success training the unconditional generation model? Any checkpoints or idea of how many epochs it takes to converge?

victor-shepardson

Hi, Is this possible to use hop-length other than 64? For example, I want to use hop length of 200 correspond to 12.5ms in 16k sampling rate audio. I tried...

cyu0913