Some publicly available TTS code repositories, that can be used out of the box.

TTS frontend repositories: Text to spectral representation

  1. ESPnet: Supports different TTS frontend algorithms

    • Tacotron2

    • Transformer-TTS

    • FastSpeech

    • FastSpeech2

    • Multispeaker Tacotron

    • Multispeaker Tacotron with Global Style Tokens (GST)

  2. NVIDIA Tacotron2

  3. r9y9 Tacotron2

Neural vocoder repositories: Maps the spectral representation output of TTS frontend to audio waveform

  1. NVIDIA wavenet

  2. r9y9 wavenet

  3. kan-bayashi wavenet

  4. NVIDIA waveglow

  5. kan-bayashi Parallel WaveGAN

  6. kan-bayashi MelGAN