raccoonML

Something else is bottlenecking performance. Since it's a cloud server the first thing I would suspect is the filesystem. Make sure the dataset is hosted on some kind of fast,...

It's possible this will be resolved with more training steps. You can also try restarting the training with a higher reduction factor to make it easier to learn attention. Once...

Which speech dataset are you using? You should be using LibriSpeech or LibriTTS if you want to compare results to the pretrained models of this repo.

The major benefit of the toolbox is the audio visualizations, in the form of speaker embeds and spectrograms. If you don't need images, a very basic interface could suffice. Maybe...

I don't intend to work on this issue, but I suggest that you come up with detailed requirements to help a developer who is interested in solving this problem. What...

Does the developer need to do anything special with wxPython to provide that accessibility info to the screen reader? Another way of stating the question is, if a wxPython interface...

Is there a way to do this with PyQT so we don't need to rewrite the interface?

A more advanced solution is to save the attention layer alignments from inference, stretch them by the desired slowdown amount, then run the decoder loop again replacing the attention network...

It can be improved by training a new vocoder model from scratch on higher quality data. You can preprocess the dataset at a higher sampling rate, and the vocoder will...

Are you using the latest code? That error message pertains to checkpoints developed for an older version of this repo, which used tensorflow.