TextToSpeech Playground - AI-Powered Text-to-Speech with Voice and Avatars

A scalable AI-powered platform for generating high-quality, natural-sounding voice from text with selectable languages and avatars.

Features

• Scalable GPU Inference: Built with Ray for distributed computing across multiple nodes and GPUs
• High-Quality Text-to-Speech: Leverages state-of-the-art models like XTTS-v2, Dia, Kokoro, Bark, and Spark-TTS
• Multiple Languages: Supports English, Arabic, Spanish, Hindi/Urdu, and more
• Avatar Selection: Choose between different voices based on gender and dialect
• RESTful API: Easy-to-use API for text input and voice file generation in MP3 format
• Automated Content Generation: Process batches of text for social media posts
• Comprehensive Monitoring: Full system metrics tracking with Ray Dashboard, Prometheus, and Grafana
• Optimized Performance: Supports asynchronous/batched TTS requests and quantized models for efficiency

Getting Started

Prerequisites

• Python 3.9+ (Python 3.13+ requires models to be custom-built)
• CUDA-compatible GPU (recommended for production)
• Docker and Docker Compose (optional, for containerized deployment)

Installation

1. Clone the repository:

   git clone https://github.com/yourusername/texttospeech-playground.git
   cd texttospeech-playground
   

2. Create a virtual environment:

   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate
   

3. Install dependencies:

   pip install -r requirements.txt
   

4. Download and prepare TTS models:

   python src/download_models.py
   

Running the Application

Local Development

uvicorn src.api.main:app --reload

Docker Deployment

docker-compose up -d

API Usage

Text-to-Speech Endpoint

POST /api/tts

Request body:

{
  "text": "Hello, this is a test message",
  "language": "en",
  "avatar": {
    "gender": "female",
    "dialect": "en-US"
  }
}

Response:

{
  "file_url": "/audio-output/2023-09-15_12-30-45_female_en-US.mp3",
  "duration_seconds": 2.5
}

Project Structure

├── audio-output/            # Generated audio files
├── models/                  # Downloaded TTS models
│   └── tts/
├── src/                     # Source code
│   ├── api/                 # API endpoints
│   ├── core/                # Core TTS functionality
│   ├── ray/                 # Ray distributed computing
│   └── monitoring/          # System monitoring
├── tests/                   # Test suite
├── Dockerfile               # Docker configuration
├── docker-compose.yml       # Docker Compose configuration
├── requirements.txt         # Python dependencies
└── README.md                # This file

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

• Hugging Face for providing access to state-of-the-art TTS models
• Ray Project for the distributed computing framework
• Coqui for their work on XTTS
• Suno for the Bark TTS model