Voice Cloner 🎤
Record 3–10 seconds of speech (or upload an audio file) and the Pocket TTS engine builds a custom voice from it on-device — no account, no upload, no API key. Preview the clone instantly, adjust pitch, speed, and volume, and save it for reuse.
Tips for a clean clone
- Use a quiet room and a decent microphone.
- Speak naturally for 3–10 seconds — a single speaker only.
- Reference clips are capped at 10 seconds and processed locally.
- Clone only your own voice or audio you have rights to use.
How it works
- Your clip is downmixed to mono and resampled to 24 kHz on-device.
- The Pocket model (146MB, downloaded once) encodes its speaker state.
- The cloned voice speaks English, German, Italian, Portuguese, or Spanish text.
💡 Memory note: the clone model (~146MB) and the design model (~305MB) load independently.
Frequently asked questions
How does browser voice cloning work here?
You record or upload 3–10 seconds of clean speech. The Pocket TTS engine runs fully in your browser: it downmixes and resamples the clip to 24 kHz, encodes the speaker identity with the mimi voice encoder, and conditions the text-to-speech model so your next generation speaks in that voice.
How long should my reference clip be?
3–10 seconds of natural speech from a single speaker works best. Clips are capped at 10 seconds and truncated if longer. A quiet room and a decent microphone matter far more than extra length.
Is my voice uploaded to a server?
No. The reference clip, the encoding, and the generated audio stay on your device. The only download is the Pocket model (~146MB, cached after the first use). Saved clones live in your browser’s IndexedDB.
Can I adjust the cloned voice after cloning?
Yes. After cloning, the adjust panel lets you move pitch (±12 semitones), speed (0.5–2×), and volume for A/B listening. An experimental trait blend can also steer the clone’s conditioning toward a built-in voice like Javert or Cosette — reset anytime to restore the true clone.
In which languages can a cloned voice speak?
The Pocket engine supports English, German, Italian, Portuguese, and Spanish bundles. The cloned voice conditions whatever language bundle is active.
Can I use a saved clone elsewhere on the site?
Yes — saved clones appear in the Pocket TTS workspace at /app/pocket/, and you can load them back here anytime. Cloning voices you do not own or have rights to use is prohibited.
Does cloning cost anything or need an API key?
No. Cloning is free, unlimited, and requires no account or API key — the model runs in your browser on your machine.