three.ws · Voice cloning

Record a short voice sample in your browser, ship it to the three.ws instant-cloning endpoint, and synthesize fresh speech with the cloned voice. For best fidelity record 20–30 seconds of clear, dry speech with consistent volume. Audio never leaves your machine until you tap Submit to clone.

1. Record a sample

00:00

2. Playback

3. Voice label

Cloned voice

No voice yet — record + submit to receive a voice_id.

Log

ready. waiting for mic permission.

Internal pipeline: MediaRecorder → POST /api/tts/eleven-clone (multipart) → ElevenLabs voices.add via the elevenlabs SDK. Instant Voice Cloning is a paid-tier feature; the free tier will return can_not_use_instant_voice_cloning.

4. Synthesize with the cloned voice

Text to speak (≤ 500 chars) Output

Sample lines

These hit the existing /api/tts/eleven endpoint with the freshly minted voice_id. Same caching + rate limit as the production TTS proxy.