Live calls
WebRTC audio carries your raw voice in parallel; the cloned-voice translation lands on top each sentence. Caller and callee can speak entirely different languages.
Products
One inbox, every language, your voice.
Call across languages. Send voice messages in your cloned voice. Text and share media in a single thread. Translation runs automatically across every channel — no toggles, no copy-paste, no mode switching.
Voice messages
Tap, talk, send. The recipient hears it in your cloned voice in their language — with the same emotion you put into the recording.
Texts
Type in your language. They read in theirs — or either side can play the text aloud in the voice tone you chose when you sent it. Replies are translated back the same way. No buttons, no mode switching, always on.
Photos & video
Share photos and videos in the same WhatsApp / MMS-style thread as your calls and texts. Today we deliver media as-is; caption translation and video re-voicing are on the roadmap.
Quality tiers
Three engines. One cloned voice.
Pick the engine bundle per call. Tiers are available on every plan — you're never locked out of higher quality by your subscription.
Fastest
Snappy, MIT-licensed engine. Best for everyday calls.
- Engine
- OpenVoice V2
- Two-stage voice cloning — keeps your timbre while base TTS handles the language
- Lightest on the GPU; great for high-volume calling
- Default for everyday calls and voice messages
Balanced
A middle gear for when you want a touch more polish.
- Engine
- Tuned OpenVoice V2
- Same engine as Fastest, with quality-side knobs turned up
- A good fit for longer calls and important voice messages
- No license restrictions — ships on every plan
Most accurate
No-compromise quality for the moments that count.
- Engine
- Chatterbox
- The warmest, most natural delivery we ship
- Hand-tuned for interviews, presentations, and broadcast-quality calls
- Available on every plan — pick it per call
The pipeline
How a sentence becomes a translated sentence — in your voice.
- 1
Real-time speech recognition
Audio streams from your phone to our servers in real time. Dedicated GPUs handle the speech-to-text transcription continuously — we don't wait for you to pause before the next stage of the pipeline picks up.
- 2
Translation
Self-hosted translation engine fluent across every language we support. Audience presets bend the register so legal stays legal and family stays warm.
- 3
Voice-cloned synthesis
The translated message lands in your cloned voice on the other side. For live calls and voice messages, the emotion in your real-time speech carries through automatically. For texts, you pick one of your five recorded tones at send time — and either side can play the text aloud. Either way, the other person hears you — not a Text-to-Speech (TTS) robot.
- 4
Delivery on every channel
Live calls ride a parallel WebRTC audio plane so you always hear something. Voice messages, texts, and media land in the same translated thread.
On the roadmap
The same words, the right register — coming next.
We're training audience-specific translation models so you'll be able to pick a register before your call or thread. Today's translations use a general-purpose model; the feedback loop — the “fix this translation” button on every call and message — trains the audiences below directly.
General
Everyday calls — neutral register, broad vocabulary.
Family
Warmer phrasing tuned for grandparents, kids, partners.
Business
Formal tone, careful with names and titles.
Tech
Preserves jargon, code names, version numbers.
Legal
Conservative phrasing for stakes-y conversations.
Medical
Patient-clinician language with precise terminology.