🎙️ Silero TTS Text-to-Speech Generator

🌟 Overview

A powerful Python-based text-to-speech generator leveraging Silero TTS models, designed for versatile and high-quality speech synthesis:

• 🌐 Multiple language support (Russian, English, German)
• 📝 Advanced SSML text processing
• 🔊 Intelligent noise reduction
• 💻 GPU/CPU compatibility
• 🎭 Flexible speaker selection
• 🌪️ Tornado-based API server for remote TTS generation

📂 Project Structure

• __main__.py: 🚀 Example script demonstrating local TTS usage
• silero_tts_processor.py: 🧠 Core TTS processor class
• tts_server.py: 🌐 Tornado-based API server for remote TTS generation
• test_request.sh: 🧪 Bash script for testing the TTS API
• requirements.txt: 📦 Project dependencies

🛠️ Prerequisites

• 🐍 Python 3.8+
• 🚀 CUDA (optional, for GPU acceleration)
• 🌐 curl for API testing (optional)
• 📊 jq for JSON parsing (optional)

🚀 Installation

1. Clone the repository:

git clone https://github.com/semyenov/silero-tts-generator.git
cd silero-tts-generator

2. Create a virtual environment:

python3 -m venv .venv
source .venv/bin/activate

3. Install dependencies:

pip install -r requirements.txt

🎬 Local Usage

from silero_tts_processor import SileroTTSProcessor

# Create TTS processor
tts = SileroTTSProcessor(
    language_id="ru",
    model_id="v4_ru",
)

# Generate speech
audio = tts.generate_speech(
    "<speak>Привет, мир</speak>",
    speaker_id="xenia",
    enhance_noise=True,
    output_filename="output.wav"
)
tts.play_audio(audio)

🌐 API Server Usage

Start the Tornado API server:

python tts_server.py

🧪 API Testing with Bash Script

A convenient bash script test_request.sh is provided to test the TTS API:

# Basic usage
./test_request.sh -t "Привет, мир"

# Advanced usage with custom parameters
./test_request.sh \
    -t "<speak>Привет, мир</speak>" \
    -s xenia

Script options:

• -t: Text to convert to speech (required)
• -s: Speaker (default: xenia)
• -h: Show help message

📄 API Endpoints

🎙️ Generate TTS

POST /tts

Request Body:

{
  "text": "<speak>Текст для синтеза речи</speak>",
  "speaker": "xenia",
  "enhance_noise": true
}

Response:

{
  "success": true,
  "filename": "generated_audio_file.wav"
}

🔍 Retrieve Audio File

GET /audio/{filename}

Retrieves the generated audio file.

🌐 Supported Languages

• Russian
• English
• German
• ...

Full list of supported languages and models can be found here.

🔧 Troubleshooting

• Ensure you have the latest version of PyTorch
• Check CUDA compatibility if using GPU
• Verify audio device settings
• Make sure curl and jq are installed for API testing

📄 License

MIT License

🤝 Contributing

Pull requests are welcome. For major changes, please open an issue first.