OpenVoice REST API

This project provides a REST API for instant voice cloning using OpenVoice V2 and MeloTTS. The API allows users to send a reference voice clip and lyrics, and returns a synthesized audio file in the cloned voice.

Project Structure

openvoice-rest-api
├── app
│   ├── __init__.py
│   ├── api.py                # Flask REST API endpoints
│   ├── models
│   │   ├── __init__.py
│   │   └── openvoice_model.py # OpenVoice V2 + MeloTTS integration
│   ├── services
│   │   ├── __init__.py
│   │   └── voice_cloning_service.py # Voice cloning and synthesis logic
│   └── utils
│       ├── __init__.py
│       └── file_utils.py     # File utility functions
├── checkpoints                # Model checkpoints (see below)
│   ├── base_speakers/ses/*.pth
│   └── converter/{checkpoint.pth, config.json}
├── Dockerfile                 # Docker build instructions
├── requirements.txt           # Python dependencies
├── config.py                  # Configuration settings
├── MeloTTS/                   # MeloTTS source (as a submodule or folder)
└── README.md                  # Project documentation

Setup Instructions

1. Clone the Repository

   git clone <repository-url>
   cd openvoice-rest-api

2. Model Checkpoints

   • Download the OpenVoice V2 checkpoints as described in the OpenVoice repo.
   • Place them in the checkpoints directory as follows:
     • checkpoints/base_speakers/ses/*.pth (e.g., en-us.pth, en-newest.pth, etc.)
     • checkpoints/converter/checkpoint.pth and checkpoints/converter/config.json

3. Build and Run with Docker

   docker build -t openvoice-rest-api .
   docker run -p 6000:6000 \
     -v $(pwd)/checkpoints:/app/checkpoints \
     -v $(pwd)/app/output:/app/app/output \
     openvoice-rest-api

   The API will be available at http://localhost:6000.

API Usage

Endpoint

• POST /clone-voice

Request (multipart/form-data)

• reference_voice: (file) The reference voice clip to clone (WAV, MP3, FLAC, OGG).
• lyrics: (string) The lyrics/text to be synthesized.
• language: (string, optional) Target language (e.g., en, es, fr, zh, jp, kr). Defaults to en.
• speaker: (string, optional) Base speaker key (e.g., en-us, en-newest). If not provided, a default is chosen.

Example (using curl)

curl -X POST http://localhost:6000/clone-voice \
  -F "reference_voice=@/path/to/voice.wav" \
  -F "lyrics=Your lyrics here" \
  -F "language=en" \
  -F "speaker=en-us"

Response

• 200 OK: JSON with audio_file_path, relative_path, and metadata.
• 400 Bad Request: If required fields are missing or invalid.
• 500 Internal Server Error: On processing errors.

Features

• Uses OpenVoice V2 for instant, multi-lingual, zero-shot voice cloning.
• Uses MeloTTS for high-quality base TTS in multiple languages.
• Reference audio is denoised and normalized for best results.
• Output audio is normalized to -1 dBFS for consistent loudness.

License

This project is licensed under the MIT License. See the LICENSE file for more details.

Maintainers

This project is actively maintained by @gbudjeakp.