Streaming and voice reference
Use Pocket TTS for streaming speech and voice references
Kyutai Pocket TTS begins returning audio while the rest of the line is still being generated. It runs on the CPU, supports English, and can condition speech on a short voice reference that you have permission to use.
Your browser keeps this engine after its first load, so later clips can work offline. Review or remove local engines on the Models page.
Choose Pocket TTS for interactive or referenced voices
Use Pocket TTS when fast first audio or a permitted voice reference is central to the project. Its 189 MB download is larger than Piper or KittenTTS, and the quality of a referenced voice depends on the recording you provide.
Start with a clean reference and a short line
- 1Use only a voice recording you own or have permission to use, with one clear speaker and little background noise.
- 2Generate a short sentence first and check whether the identity, clarity, and pace remain stable.
- 3Move to longer sections only after the reference and written script produce a consistent result.
Specifications
| Developer | Kyutai |
|---|---|
| Parameters | 100 million |
| Download size | 189 MB (int8) |
| Output | 24.0 kHz mono |
| Browser voices | 10 |
| Languages | English |
| Voice cloning | Yes |
| Streaming | Yes — audio starts before generation finishes |
| License | CC-BY-4.0 (weights), MIT (code) |
| Speed | Begins producing audio during generation on a current multi-core CPU. |
| Runtime | WebAssembly (CPU) |
Specifications checked 2026-07-21 against:
Explore related downloads
The browser studio and downloadable catalog are separate. Use the catalog to find independently published model files and check each source before downloading.
Good fit when
- +Streaming playback during generation
- +Supports short voice-reference recordings
- +Runs locally on a CPU
Before you choose it
- −English only in the current browser implementation
- −189 MB initial download
- −Reference quality depends on the source recording
Voices (10)
See it alongside another model
Questions about Kyutai Pocket TTS
Does AIVoices charge to use Kyutai Pocket TTS?
AIVoices does not charge generation credits for this browser model. The listed model license is CC-BY-4.0 (weights), MIT (code). Check the upstream model and voice terms before use.
What hardware does Kyutai Pocket TTS need?
A current desktop browser and CPU are enough for this model. The 189 MB model is stored in your browser after the first download.
Is my text private when using Kyutai Pocket TTS here?
The supported model executes inside your browser tab. AIVoices does not send the text to a server for speech generation.
Can I use Kyutai Pocket TTS output commercially?
Commercial use depends on the model, selected voice, and source-data terms. This page lists CC-BY-4.0 (weights), MIT (code) for the model, but you should still review the current upstream terms.