how to use squeezewave with tts to generate voice

[tianrengao]

how to use squeezewave with tts to generate voice

[tianrengao]

Hi, there are actually two parts in the TTS pipeline. The front-end part generates high quality mel-spectrogram from text, and the back-end part generates real audio given high quality mel-spectrogram. Basically, to generate voice from text using a whole TTS pipeline, you will need both the front-end and the back-end.
For the front-end part, some models like tacotron2 and fastspeech do very good job. And SqueezeWave is designed to be an efficient model for the back-end part. There are also some other models designed for the back-end part, such as waveRNN, waveGlow, and LPCNet.
So here, to generate voice, we assume you already have high quality mel-spectrogram as input for SqueezeWave.

[li-xx-5]

@tianrengao hello,may i ask how to use the denoiser.py .thank you

[li-xx-5]

i cannot run that.