The call comes in
It reaches a line that answers with an interpreter. The caller hears a notice in their language: the call is translated automatically, and short sentences work best.
Your customer calls speaking German, French or Japanese, and your team answers in English. Throughout the call, each person hears the translation of what the other says: never their own voice translated, never the other person's original voice.
There isn't a single interpreter trying to pull apart two mixed voices. There are two, and each works in one direction only: one listens only to your customer and speaks to your team; the other listens only to your team and speaks to your customer.
Your customer hears what your team says, already in their language; your team hears what the customer says, already in theirs. Nobody hears their own voice translated or the other person's original voice.
The listening rules live in the call itself, not in the interpreter. Even if an interpreter makes a mistake or drops out, it can't make anyone hear what isn't meant for them.
References, amounts, names and customer numbers pass through untouched: "my customer number is 456" arrives complete, digit by digit.
Each interpreter writes down what was said and its translation. That's where the live captions and the bilingual transcript come from.
From the moment the phone rings until someone hangs up, this is how your customer and your team experience it.
It reaches a line that answers with an interpreter. The caller hears a notice in their language: the call is translated automatically, and short sentences work best.
Before picking up, the softphone already shows that the call comes with a translator, and between which two languages.
In under a second the call moves to a private room with your customer, your agent and the two interpreters. Everyone joins muted until it's settled who hears whom.
The interpreters take a second or two to join, and the notice covers that gap. Once they're in, your agent hears a short tone that means "go ahead".
Every sentence is translated continuously as the conversation moves on, keeping the speaker's intonation and pace. The captions appear next to the phone.
The room closes for everyone. The bilingual transcript and the recordings are saved, and the translation time used is closed off.
The translation runs a few seconds behind the speaker, as it would with a simultaneous interpreter. That's why the softphone tells your agent, at every moment, whether it's their turn to speak or to wait, and puts the captions of the conversation in front of them.
The caller knows from the start that they're speaking through an AI interpreter, as European transparency rules require. And while the translation is on its way, they hear soft music, so they know the call is still there.
Your customer can speak any of more than 70 languages: theirs is recognised from their first sentence and the interpreter answers them in it. Your team answers in its own language, which you pick for each line from these 16.
With "detect automatically", the caller's first sentence is enough to know their language —out of more than 70—, and the interpreter speaking to them adjusts there and then.
If your agent greets the customer before they say anything, what they say is held —for up to 20 seconds— and translated in full as soon as the language is known.
A line just for German customers speaks German from the first second. The fixed language overrides the detected one, so similar languages, like Spanish and Portuguese, never get mixed up.
Your agent can change the customer's language from the softphone, and the last few seconds are translated again into the right language.
If the first sentence already comes in your team's language, the call switches to direct with no delay and the softphone says why. Only the first sentence counts: a "thank you" mid-call doesn't turn translation off.
And your customer, in any of 70+: theirs is recognised automatically.
Whatever happens to the translation, your customer and your agent end up hearing each other. And the softphone says what happened.
After a drop, your agent presses "Retry" and fresh interpreters join the same call.
While the call is live, supervisors follow it with live captions. When it ends, the record keeps the full bilingual transcript: who said what, in which language, and what the other person heard.
You choose which lines answer with an interpreter, which team takes the calls and in which language. Everything else about the line stays the same: its opening hours, what happens if nobody picks up —voicemail, another phone, a message— and out-of-hours.
Usage runs while translation is on. Hold time, direct time and whatever happens after a drop don't count.
The AI interpreter is an add-on, available on any plan.
See plansYour customer can speak any of more than 70 languages, recognised automatically from their first sentence. Your team answers in its own, which you pick for each line from 16: Spanish, English, German, French, Italian, Portuguese, Dutch, Catalan, Basque, Galician, Polish, Romanian, Russian, Chinese, Arabic and Japanese.
No. With "detect automatically", the language is recognised from their first sentence, and your agent's greeting is held so it can be translated in full. If a line is for one language only, you fix it and translation starts from the first second. And your agent can always correct it mid-call.
Through a notice in their language, before the queue or on pick-up. European AI transparency rules require it, so it always plays; you choose whether it's the standard text, one you write or your own recording.
Yes. With the softphone switch, translation goes off and you talk directly, instantly. And if the customer already speaks your team's language, the call switches to direct by itself.
The full bilingual transcript on the record —who said what, in which language and what the other person heard—, your customer's recording with what they said and heard, and a separate one with your agent's original voice.
Only time with translation on counts. Hold time, direct time and whatever happens after a drop are left out.
No. Your team answers from the softphone, in the browser, and translation is switched on in each line's settings.
A straight conversation about your volume, your channels and where to start. No generic proposal.