/Multispeaker-Text-To-Speech-Synthesis-Web-App

Primary LanguagePython

Text to Speech Voice Cloning

Reference from https://github.com/CorentinJ/Real-Time-Voice-Cloning

System Dependencies

  • docker version: 18.09.3, build 774a1f4
  • Docker Engine version: 18.09.3, build 774a1f4
  • docker-compose version: 1.24.0, build 0aa59064
  • nvidia-docker version: 2.0.3
  • CUDA version: >= 10.0

Requirements

Need at least 6GB GPU to run

Setup

  1. Create your own copy of .env by running cp .env_sample .env
  2. docker-compose up --build

API call

curl -X POST -F "files[]=@full.wav" -F "text=They developed superpowers after years of drinking from a lead-poisoned water supply But just having incredible abilities doesn't make them superheroes" localhost:8080/api/tts -o r.json

Limitations

  • Can only be able to generate one sentence at a time (For 6GB GPU)
  • Lack of crash handling (GPU OOM)

Future Work