waveglow-vocoder 0.1.0

A vocoder that can convert audio to Mel-Spectrogram and reverse with WaveGlow, all on GPU(if avaliable).

WaveGlow Vocoder

A vocoder that can convert audio to Mel-Spectrogram and reverse with WaveGlow, all on GPU(if avaliable).
Most code are extracted from Tacotron2 and WaveGlow of Nvidia.

Install

pip install waveglow-vocoder

Example

Performance

CPU(Intel i5):

GPU(GTX 1080Ti):

Usage

Load wav file as usual

import librosa

y,sr = librosa.load(librosa.util.example_audio_file(), sr=22050, mono=True, duration=10, offset=30)
y_tensor = torch.from_numpy(y).to(device='cuda', dtype=torch.float32)

Apply mel transform, this would be done on GPU if avaliable.

from waveglow_vocoder import WaveGlowVocoder

WV = WaveGlowVocoder()
mel = WV.wav2mel(y_tensor)

Decoder it with Waveglow.

NOTE:
As the parameter of pre-trained model is alignment with Tacotron2, one might get totally noise if the Mel spectrogram comes other function than wav2mel(an alias for TacotronSTFT.mel_spectrogram).
Support for librosa and torchaudio is under development.

wav = WV.mel2wav(mel)

Other pretrained model / Train with your own data

This vocoder will download pre-trained model from pytorch hub on the first time of initialize.
You can also download the latest model from WaveGlow, or with your own data and pass the path to the waveglow vocoder.

config_path = "your_config_of_model_training.json"
waveglow_path="your_model_path.pt"
WV = WaveGlowVocoder(waveglow_path=waveglow_path, config_path=config_path)

Then use it as usual.

TODO

• pip
• WaveRNN Vocoder
• MelGAN Vocoder
• examples
• performance
• support librosa Mel input
• CPU support

Reference

• WaveGlow
• Tacotron2
• Wavenet Vocoder