This code is based on NVIDIA implementation of Tacotron2 and Waveglow. It is shaved to reduce dependencies and simplify code

https://github.com/NVIDIA/tacotron2 https://github.com/NVIDIA/waveglow

Hardware requirements

For realtime synthesis it requires GPU card like GTX 1080 or RTX 2080

Models

Download model here ~900Mb

http://alphacephei.com/tts/tacotron-models.tar.gz

Unpack them into folder tacotron-models here.

Usage

Unpack models to models folder and run test.py, it will create the audio file test.wav