AI Phone Interpreter Now Speaks 31 Languages — Any Two, Directly

Our live AI phone interpreter used to speak 8 languages. As of today it speaks 31 — and, more importantly, any two of them pair directly on the same call. A Japanese speaker and an Arabic speaker can talk normally now; the interpreting runs between their two languages, not through English as a bridge.

The price did not change.

The 23 languages we added

The original eight were Turkish, English, German, French, Spanish, Italian, Arabic and Russian. Today we added:

Chinese (Mandarin), Japanese, Korean, Hindi, Portuguese, Dutch, Polish, Swedish, Ukrainian, Indonesian, Greek, Romanian, Czech, Danish, Finnish, Norwegian, Bulgarian, Hungarian, Slovak, Vietnamese, Malay, Tamil and Croatian.

If you searched for a Chinese interpreter, a Japanese interpreter, a Vietnamese interpreter or a Korean interpreter and landed here — those all work now, live, on a phone call or a shareable browser link.

What actually changed (the honest version)

Speech-to-text was never the limit. Deepgram Nova-3 already transcribes 100+ languages. The ceiling is the speech side — the voice that talks back. ElevenLabs Flash v2.5 covers 32 languages, and the interpreter supports the intersection that runs cleanly in both directions: 31. We would rather ship 31 that sound right than claim 100 we cannot speak back.

Two things fall out of that decision:

• Every system announcement is localized too. Greeting, "please hold", "you're connected", and "no answer" each play in the participant's own language. The person waiting on the line hears why in a language they understand — which is the difference between a call that feels alive and one that feels dropped.
• Any pair, not a hub. Because both directions are first-class, you are not limited to "X to English". Mandarin↔Arabic, Japanese↔Russian, Vietnamese↔ German — all direct.

AI interpreter vs. a human interpreter

To be clear about the last row: for court, certified medical, or legally binding work, book a licensed human interpreter. This is built for the enormous number of everyday calls that get no interpreter today because booking one is too slow or too expensive — a patient scheduling surgery, a supplier, a driver, a tenant, a guest at the front desk.

Which languages are real demand — and where

We pulled live search volumes so this isn't a guess. In the US market, the strongest pull among the new languages is Chinese/Mandarin (multiple "chinese interpreter" / "mandarin interpreter" queries around 390/mo each), then Vietnamese and Japanese interpreter (~390 and ~210/mo), Korean (~260/mo) and Portuguese (~210/mo). There's a steady medical-interpreter tail across all of them — "vietnamese medical interpreter", "korean medical interpreter", "chinese medical interpreter" — which maps exactly to clinics and hospitals serving immigrant and medical-travel patients.

Hindi searches low despite the speaker base (Hindi speakers often default to English on business calls), so we'd point Indian-market value at business and support use, not interpreter search intent.

No new pricing, no new setup

Same interpreter, same shareable link, same wallet. If you already use it, the 23 new languages are just there — pick any two and talk. If you don't yet, it's the same $0.25/min interpreter it was yesterday, now with four times the language coverage.