Kyutai TTS

What Kyutai TTS is

Kyutai TTS is a free, open-source text-to-speech offering from Kyutai, an open-science AI research lab based in Paris. It includes two models: Pocket TTS and Kyutai TTS 1.6B. Pocket TTS has 100 million parameters and is designed to generate speech in real time on a CPU. Its listed languages are English, French, German, Portuguese, Italian, Spanish, and Dutch. The model provides Python, command-line, and local HTTP-server interfaces, and a browser demo lets visitors try it without installing software. Kyutai TTS 1.6B can begin generating audio before the full text arrives, a streaming approach intended for low-latency voice assistants. Kyutai says this model is used in Unmute, its interactive real-time speech demo. The offering supports voice cloning and API access, with WAV as its listed export format. Pocket TTS requires Python 3.10 through 3.14 and PyTorch 2.5 or later, though it does not require the GPU version of PyTorch. Its repository reports about 200 milliseconds to the first audio chunk and roughly six times real-time generation on a MacBook Air M4 CPU. The Pocket TTS repository identifies its license as MIT.

Who it suits

Pocket TTS suits developers who want a CPU-oriented speech model with Python, command-line, or local HTTP-server access, as well as people who want to try it in a browser. Kyutai TTS 1.6B is aimed at server and low-latency applications such as voice assistants that need to start speaking before all input text is available.

Getting started

1. Open the Kyutai TTS website and try Pocket TTS in its browser demo without installing anything.
2. Choose Pocket TTS or Kyutai TTS 1.6B based on whether you need CPU-oriented generation or streaming for low-latency applications.
3. For Pocket TTS, use Python 3.10 through 3.14 and install PyTorch 2.5 or later; the GPU version is not required.
4. Use Pocket TTS through its Python API, command-line interface, or local HTTP server.
5. Use the API access or platform environment that fits your intended setup.

Verdict

Choose Kyutai TTS if you want free, open-source speech generation, particularly Pocket TTS for CPU use or TTS 1.6B for streaming applications. Look elsewhere if your Pocket TTS workflow depends on creating pauses by inserting silence in the input text.