Mael

I'm not sure which command-line tool it is, could you show me the command you are using?

Using a K80 GPU and batch size of 32 I have about 0.8step/s, and the GPU seems to be around 90% utilisation

I'm using `sample_rate=22050` so perhaps this could explain some difference in the speed

I'm using the GPU, it shows the use as close to 90% so I think it is working

And python3.7 and pytorch 1.0

I'm not completely familiar with Festvox build script, I had the idea that MaryTTS voice building tools were perhaps easier to use, isn't it the case? and are there well-documented...

Thanks @psibre very clear, I will take a look at this

There is a similar issue when training on LibriTTS

I'm training on the real mel-spectrogram, and without global conditioning, no GTA

How should it be set?