Download
Parallel WaveGAN is an unofficial PyTorch implementation of several state-of-the-art non-autoregressive neural vocoders, centered on Parallel WaveGAN but also including MelGAN, Multiband-MelGAN, HiFi-GAN, and StyleMelGAN. Its main goal is to provide a real-time neural vocoder that can turn mel spectrograms into high-quality speech audio efficiently. The repository is designed to work hand-in-hand with ESPnet-TTS and NVIDIA Tacotron2-style front ends, so you can build complete TTS or singing voice synthesis pipelines. It includes a large collection of “Kaldi-style” recipes for many datasets such as LJSpeech, LibriTTS, VCTK, JSUT, CMU Arctic, and multiple singing voice corpora in Japanese, Mandarin, Korean, and more. The project provides pre-trained models, Colab demos, and example configurations, allowing researchers to quickly evaluate vocoder quality or adapt models to new datasets.
Features
- PyTorch implementations of Parallel WaveGAN, MelGAN, Multiband-MelGAN, HiFi-GAN, and StyleMelGAN
- Real-time neural vocoder compatible with ESPnet-TTS and Tacotron2 front ends
- Extensive set of Kaldi-style recipes for speech and singing datasets in multiple languages
- Pretrained models and Colab demos for quick listening tests and prototyping
- Flexible training pipeline with support for multi-GPU and distributed setups
- Very low real-time factor for fast mel-to-waveform conversion suitable for deployment
Project Samples
