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.
pip install neon-tts-plugin-glados
"tts": {
"module": "neon-tts-plugin-glados"
}
A docker container using ovos-tts-server is available
You can build and run it locally
docker build . -t gladostts
docker run -p 8080:9666 gladostts
use it http://localhost:8080/synthesize/hello
Or use a public instance: