multispeaker-glow-tts

Tensorflow Implementation of Glow-TTS, Jaehyeon Kim et al., in NeurIPS 2020.

• Glow-TTS: A Generative Flow for Text-to-Speech via Monotonic Alignment Search [arXiv:2005.11129]
• Full code is based on original github repository jaywalnut310/glow-tts

Structure

Training

Inference

Requirements

Tested in python 3.7.11 conda environment, requirements.txt

Usage

Put dataset specific symbols list in config file base_blank.json

train, valid filelists format

absolute_filepath|numerical_speaker_id|transcript

To start training, run train_ddi.sh

To inference the audio, run inference.py