Voice Cloning using OuteTTS

A Flask-based API for voice cloning and text-to-speech generation using the OuteTTS library.

IMPORTANT: This application requires a GPU (NVIDIA L4 or greater recommended) for optimal performance. Voice cloning is computationally intensive and will be significantly slower without adequate GPU resources.

Overview

This project enables you to:

• Generate voice maps from audio samples
• List available voice maps
• Generate speech using cloned voices with custom text

The system uses Google Cloud Storage for storing voice maps and generated audio files, making them accessible via signed URLs.

Directory Structure

└── hrushik-reddy-voice-clone/
    ├── README.md
    ├── api.py
    ├── postman.json
    └── requirements.txt

Features

• Voice Map Generation: Create voice fingerprints from audio samples
• Voice Cloning: Generate speech with cloned voices using custom text
• Cloud Storage: Store and retrieve voice maps and generated audio using Google Cloud Storage
• API Interface: Easy-to-use REST API endpoints

Requirements

• Python 3.8+
• NVIDIA GPU (L4 or greater recommended)
• CUDA and cuDNN properly installed
• Google Cloud credentials
• OuteTTS library
• Flask and additional dependencies (see requirements.txt)

Installation

1. Clone the repository:

git clone https://github.com/hrushik-reddy/voice-clone.git
cd hrushik-reddy-voice-clone

2. Install dependencies:

pip install -r requirements.txt

3. Set up Google Cloud credentials:

export GOOGLE_APPLICATION_CREDENTIALS="path/to/your/service-account-file.json"

Alternatively, place your credentials file at var/secrets/google/proposals-creds.json

4. Create necessary directories:

mkdir -p voice_maps

Usage

Running the API Server

python api.py

The server will start on http://0.0.0.0:5000

API Endpoints

1. Generate Voice Map

Create a voice map from an audio sample:

POST /generate_voicemap

Request Body:

{
  "signed_url": "https://storage.googleapis.com/your-audio-file.mp3",
  "speaker_name": "custom_speaker_name"
}

Response:

{
  "status": "success",
  "speaker_name": "custom_speaker_name",
  "voice_map_path": "voice_maps/custom_speaker_name_voice_map.json",
  "voice_map_url": "https://storage.googleapis.com/...",
  "time_taken": "2.45 s"
}

2. List Voice Maps

Get a list of all available voice maps:

GET /list_voice_maps

Response:

{
  "status": "success",
  "voice_maps": [
    {
      "speaker_name": "custom_speaker_name",
      "file_path": "voice_maps/custom_speaker_name_voice_map.json"
    }
  ]
}

3. Generate Voice

Generate speech using a cloned voice:

POST /generate_voice

Request Body:

{
  "speaker_name": "custom_speaker_name",
  "text": "Hello, this is a test of voice generation."
}

Response:

{
  "status": "success",
  "speaker_name": "custom_speaker_name",
  "audio_url": "https://storage.googleapis.com/...",
  "time_taken": "1.23 s"
}

Postman Collection

A Postman collection is included in the repository (postman.json) to help you test the API endpoints. Import this collection into Postman to get started quickly.

Technical Details

• Audio Processing: Uses PyDub to process and normalize audio files
• Voice Cloning: Leverages OuteTTS for voice cloning and speech generation
• Storage: Uses Google Cloud Storage for cloud-based file storage
• API Framework: Built with Flask for lightweight API endpoints

Limitations

• Audio samples are trimmed to 20 seconds for voice map generation
• Requires proper Google Cloud credentials with access to the specified bucket

License

MIT License

Author

Hrushik Reddy