Sarvam TTS

sarvam-tts is a Python client for the Sarvam AI text-to-speech API. It provides a clean SarvamTTS interface for multilingual Indian language synthesis, configurable output settings, and chunked request handling for reliable audio generation.

⚠️ This project is provided for educational and research purposes only. Please comply with Sarvam AI's terms of service and acceptable use policies when using the API.

✨ Features

• Multi-language TTS support for Indian languages
• Simple SarvamTTS API for generating MP3 output
• Configurable TTSConfig for language, voice, pace, temperature, sample rate, and timeout
• Concurrent chunk generation for long texts
• Package metadata and install support via setup.py

📁 Project Structure

sarvam-tts/
├── README.md
├── LICENSE
├── requirements.txt
├── setup.py
├── example.py
├── sarvam_tts/
│   ├── __init__.py
│   ├── config.py
│   ├── exceptions.py
│   ├── utils.py
│   └── providers/
│       ├── __init__.py
│       ├── base.py
└───────└── sarvam.py

Package contents

• sarvam_tts/__init__.py exports the public API: SarvamTTS, TTSConfig, TTSException, and TTSRequestError
• sarvam_tts/config.py contains supported languages, default API URL, and configuration validation
• sarvam_tts/providers/sarvam.py implements the Sarvam AI request flow and audio saving logic
• sarvam_tts/providers/base.py defines the provider abstraction
• sarvam_tts/utils.py contains the multilingual sentence tokenizer
• example.py demonstrates basic usage

🛠️ Installation

Using PyPI (Recommended)

pip install sarvam-tts

Clone Locally

git clone https://github.com/sujalrajpoot/sarvam-tts.git
cd sarvam-tts
pip install -r requirements.txt

🧪 Quick Start

from sarvam_tts import SarvamTTS

client = SarvamTTS()

output_path = client.tts(
    text="Hello from Sarvam TTS!",
    language="english",
    voice="shreya",
    pace=1.0,
    temperature=0.6,
    sample_rate=22050,
    output_filepath="output.mp3",
)

print(output_path)

You can also customize the configuration directly:

from sarvam_tts import TTSConfig, SarvamTTS

config = TTSConfig(
    language="hindi",
    voice="shreya",
    pace=1.0,
    temperature=0.4,
    sample_rate=22050,
    output_path="demo/output.mp3",
)

client = SarvamTTS(config=config)
client.tts("नमस्ते, यह एक उदाहरण है।")

🧠 Supported Languages

• english
• hindi
• bengali
• tamil
• telugu
• kannada
• malayalam
• marathi
• gujarati
• punjabi
• odia

🎤 Example Voice Names

The library supports common voice names such as:

• shreya
• shubh
• manan
• ishita
• priya
• suhani
• ashutosh
• ritu
• amit
• sumit
• pooja
• simran
• rahul
• kavya
• ratan
• shruti
• aditya
• soham
• rehan
• vijay
• tarun
• anand
• aayan
• rohan
• dev
• sunny
• kabir
• varun
• neha
• mani
• mohit
• rupali
• advait
• roopa
• tanya
• gokul
• kavitha

🔧 API Reference

SarvamTTS.tts(...)

Parameters:

• text – text to synthesize
• language – optional override for the language
• voice – optional override for the voice
• pace – optional override for speaking pace
• temperature – optional override for voice variability
• sample_rate – optional override for sample rate
• output_filepath – optional output path

TTSConfig

The configuration object provides defaults for:

• timeout
• verbose
• output_path
• language
• voice
• pace
• temperature
• sample_rate
• api_url

🖼️ How It Works

1. SarvamTTS accepts text and configuration values.
2. The text is split into sentence-level chunks.
3. Each chunk is sent to the Sarvam AI endpoint concurrently.
4. Audio responses are merged into a single MP3 payload.
5. The final audio is saved to the requested output path.

🧪 Running the Example

python example.py

The script generates sample audio files in a Test/ directory.

🤝 Contributing

Contributions are welcome. To contribute:

1. Fork the repository
2. Create a feature branch
3. Commit your changes
4. Open a pull request

📄 License

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