Description

OVOS TTS plugin for GladosTTS

The initial, regular Tacotron model was trained first on LJSpeech, and then on a heavily modified version of the Ellen McClain dataset (all non-Portal 2 voice lines removed, punctuation added). The Forward Tacotron model was only trained on about 600 voice lines. The HiFiGAN model was generated through transfer learning from the sample. All models have been optimized and quantized.

Install

pip install neon-tts-plugin-glados

Configuration

  "tts": {
    "module": "neon-tts-plugin-glados"
    }

Docker

build it

docker build . -t gladostts

run it

docker run -p 8080:9666 gladostts

use it http://localhost:8080/synthesize/hello