🎙️Pocket TTS Voice
Kyutai AI voices in 5 languages, on-device. Free, private, no signup — files never leave your browser.
Pocket TTS by Kyutai (100M params, built-in preset voices) — models download once, then run on-device. Custom voice cloning is disabled to prevent misuse. Web runtime adapted from KevinAHM's demo (Apache-2.0).
Many voices. Five languages. Zero servers.
Kyutai Pocket TTS with built-in voices, streaming from your own device. No upload, no signup, no watermark. Cloning disabled to prevent misuse.
How to use Pocket TTS Voice
Make preset voices say anything. Pick one of 8 built-in voices per language — English, German, Italian, Portuguese or Spanish — and generate natural speech with streaming low-latency synthesis. Models download once and are cached; everything after that is instant and private. Custom voice cloning is disabled to prevent misuse.
- 1Type or paste up to 2000 characters.
- 2Pick a language and a built-in voice.
- 3Hit Generate and download the WAV voice-over.
Use cases
What pocket tts voice does
- ✓8 built-in voices per language, no uploads needed.
- ✓Five languages: English, German, Italian, Portuguese, Spanish.
- ✓Streaming synthesis with ~200ms time to first audio.
- ✓High-quality 24 kHz WAV download for editors.
- ✓Fully local ONNX: no uploads, no accounts, no watermarks.
Specs
- Speed
- ~6x real-time on CPU, streaming
- Input
- Text up to 2000 characters
- Output
- WAV voice-over (24 kHz)
- Privacy
- On-device ONNX, zero uploads
- Cost
- Free, unlimited, no signup
Frequently asked questions
Is my text sent to a server?
No. Pocket TTS runs 100% in your browser via ONNX. Your text never leaves your device.
Which languages and voices are available?
Five languages (English, German, Italian, Portuguese, Spanish), each with 8 built-in voices.
Why does the first run take longer?
The voice models download once and are cached by your browser. Every run after that starts faster.
What format is the download?
A 24 kHz WAV voice-over file you can drop into any video or podcast editor.