This code is forked from https://github.com/fatchord/WaveRNN and optimized for Mozilla-TTS.
Models | Dataset | Commit | Audio Sample | TTS model | Details |
---|---|---|---|---|---|
mold model | LJspeech | 8a1c152 | soundcloud | Tacotron2-iter-260K | Model with Mixture of Logistic Distribution |
10bit model | LJSpeech | faea90b | soundcloud | Tacotron2-iter-260K | 10bit Softmax output |
universal vocoder | LibriTTS | 12c8744 | soundcloud | - | (details) |
Check this TTS notebook to see TTS+WaveRNN in action. To train your own model, you can use ExtractTTSSpectrogram to generate spectrograms by TTS and train WaveRNN. It might be also interesting to check this TTS issue to catchup with the current state.