fedirz / parler-tts-server

Parler-TTS-Server

This repository provides a server with an OpenAI compatible API interface for Parler-TTS.

Quick Start

Docker

docker run --detach --volume ~/.cache/huggingface:/root/.cache/huggingface --publish 8000:8000 fedirz/parler-tts-server

Using a fine-tuned model. See main.py for configurable options

docker run --detach --volume ~/.cache/huggingface:/root/.cache/huggingface --publish 8000:8000 --env MODEL="parler-tts/parler-tts-mini-expresso" fedirz/parler-tts-server

Docker Compose

curl -sO https://raw.githubusercontent.com/fedirz/parler-tts-server/master/compose.yaml
docker compose up --detach parler-tts-server

Usage

Saving to file

curl -s -H "content-type: application/json" localhost:8000/v1/audio/speech -d '{"input": "Hey, how are you?"}' -o audio.mp3

Specifying a different format.

curl -s -H "content-type: application/json" localhost:8000/v1/audio/speech -d '{"input": "Hey, how are you?", "response_type": "wav"}' -o audio.wav

Playing back the audio

curl -s -H "content-type: application/json" localhost:8000/v1/audio/speech -d '{"input": "Hey, how are you?"}' | ffplay -hide_banner -autoexit -nodisp -loglevel quiet -

Describing the voice the model should output

curl -s -H "content-type: application/json" localhost:8000/v1/audio/speech -d '{"input": "Hey, how are you?", "voice": "Feminine, speedy, and cheerfull"}' | ffplay -hide_banner -autoexit -nodisp -loglevel quiet -

OpenAI SDK usage example can be found here

Citations

@misc{lacombe-etal-2024-parler-tts,
  author = {Yoach Lacombe and Vaibhav Srivastav and Sanchit Gandhi},
  title = {Parler-TTS},
  year = {2024},
  publisher = {GitHub},
  journal = {GitHub repository},
  howpublished = {\url{https://github.com/huggingface/parler-tts}}
}

@misc{lyth2024natural,
      title={Natural language guidance of high-fidelity text-to-speech with synthetic annotations},
      author={Dan Lyth and Simon King},
      year={2024},
      eprint={2402.01912},
      archivePrefix={arXiv},
      primaryClass={cs.SD}
}