Kitten TTS is a lightweight text-to-speech library built on ONNX that synthesizes speech from text using compact neural models.
The tool addresses the need for high-quality voice synthesis on resource-constrained devices by packaging state-of-the-art TTS models into extremely small footprints, ranging from 15 million to 80 million parameters and occupying 25 to 80 megabytes on disk. It runs efficiently on CPU without requiring GPU acceleration. The library includes eight built-in voices and offers adjustable speech speed control, text preprocessing for handling numbers and currencies, and 24 kHz audio output quality.
Developers working on edge deployment, mobile applications, or systems with limited computational resources should consider this tool. It suits projects where model size and CPU efficiency are critical constraints. The project is in developer preview status, meaning APIs may change between releases. Commercial support is available for integration assistance, custom voices, and enterprise licensing.
The project maintains a substantial base of adopters who report real-world issues, with almost all open issues raised by outside users rather than the core team. Maintainers typically respond to new issues and pull requests within a day.
