Kyutai TTS
Kyutai TTS is a free, open-source text-to-speech offering with two models: Pocket TTS and Kyutai TTS 1.6B.
What Kyutai TTS is
Kyutai TTS is a free, open-source text-to-speech offering with two models: Pocket TTS and Kyutai TTS 1.6B. Pocket TTS has 100 million parameters and is designed to run in real time on a CPU. It supports English, French, German, Portuguese, Italian, Spanish, and Dutch, and provides a Python API, command-line interface, and local HTTP server. A browser demo is available without installation. Kyutai TTS 1.6B can start generating audio before the full text input arrives, a design intended for low-latency voice assistants; Kyutai says it is used in the Unmute interactive real-time speech demo. The offering supports voice cloning and WAV export. Pocket TTS requires Python 3.10 through 3.14 and PyTorch 2.5 or later, but not the GPU version of PyTorch. Its repository lists silence in input text as an unsupported way to create pauses. The Pocket TTS license is MIT. Both listed plans are free: Pocket TTS is an open-source model, while TTS 1.6B is an open-source streaming model designed for servers and low-latency applications.
Good
- Both listed plans are free
- Pocket TTS is designed for CPU use
- Seven listed language options
- Browser demo requires no installation
Watch out
- Pocket TTS requires Python 3.10–3.14
- Requires PyTorch 2.5 or later
- Input text cannot add pauses through silence
Who it suits
Kyutai TTS suits developers seeking open-source speech generation, including CPU-focused use with Pocket TTS or streaming applications with TTS 1.6B. Its seven listed languages and voice cloning may also fit users who need those capabilities.
Verdict
The two models offer distinct approaches: Pocket TTS targets CPU use, while TTS 1.6B streams before all text arrives. Review Pocket TTS’s runtime requirements and pause limitation before choosing it.
Plans and prices
From Kyutai TTS's own pricing page, checked 12 Oct 2001.
| Plan | Price | What it gives you |
|---|---|---|
| Kyutai Pocket TTS | Free | Open-source model; 100M parameters; runs on CPU; voice cloning; seven listed languages |
| Kyutai TTS 1.6B | Free | Open-source streaming model; designed for servers and low-latency applications |
The facts
Each one read from Kyutai TTS's own pages, with the date it was checked.
What it does
- API access
- Yeschecked 20 Sep 2001
- Export formats
- WAVchecked 20 Sep 2001
- Integration
- Kyutai says TTS 1.6B is used in Unmute, its interactive real-time speech demo.checked 12 Oct 2001
- Interfaces
- Pocket TTS provides a Python API, command-line interface, and local HTTP server.checked 12 Oct 2001
- Languages
- The Pocket TTS repository lists English, French, German, Portuguese, Italian, Spanish, and Dutch support.checked 12 Oct 2001
- License
- The Pocket TTS GitHub repository identifies its license as MIT.checked 12 Oct 2001
- Limit
- The Pocket TTS repository lists adding silence in the input text to create pauses as unsupported.checked 12 Oct 2001
- Organization
- Kyutai describes itself as an open-science AI research lab based in Paris.checked 12 Oct 2001
- Performance
- The repository reports about 200 ms to the first audio chunk and roughly six times real-time generation on a MacBook Air M4 CPU.checked 12 Oct 2001
- Pocket TTS
- Pocket TTS is a 100-million-parameter model designed to run in real time on a CPU.checked 12 Oct 2001
- Product
- Kyutai describes TTS as two open-source text-to-speech models: Pocket TTS and Kyutai TTS 1.6B.checked 12 Oct 2001
- Runtime requirements
- Pocket TTS supports Python 3.10 through 3.14 and requires PyTorch 2.5 or later; it does not require the GPU version of PyTorch.checked 12 Oct 2001
- Streaming model
- Kyutai TTS 1.6B can begin generating audio before the full text input is available, a design the maker says suits low-latency voice assistants.checked 12 Oct 2001
- Voice cloning
- Pocket TTS can clone a voice from an audio sample when run locally.checked 12 Oct 2001
Where it runs
- Browser demo
- The maker says visitors can try Pocket TTS in a browser without installing anything.checked 12 Oct 2001
- Platforms
- web,macos,ios,api,self_hostedchecked 20 Sep 2001
The company
- Founded
- 2023checked 28 Sep 2001
- Headquarters
- Paris, Francechecked 28 Sep 2001
Alternatives to Kyutai TTS
- 1
Voicebox
Free planOpen sourceAndroidiOSWebAPI8.5From $1 a month - 2
iSpeech
Free planOpen sourceAndroidiOSWebAPI8.4$50, paid once - 3
Resemble AI
Free planAndroidiOSWebAPI8.3From $280 a month - 4
Voice.ai
Free planAndroidiOSWebAPI8.2From $5 a month - 5
eSpeak NG
Free planOpen sourceAndroidiOSAPILinux7.8Free plan - 6
FineVoice AI Voice Changer
Free planWebAPImacOSWindows7.8From $6 a month