Most "multilingual" video calls are not multilingual at all. They are English calls with a few people struggling. The native speakers set the pace, the rest listen harder than they talk, and the best idea in the room stays in the head of the person who did not want to risk saying it badly.

There is a second common pattern: one language on the call, one person quietly relaying to a colleague, everything running at half speed. It works for a status update. It fails for a negotiation, a workshop or a Q&A.

Simultaneous interpretation was built for exactly this, but for decades it required a booth, two interpreters per language and a budget that only the UN and EU institutions could sustain for routine meetings. AI interpretation changes the economics, and a meeting room designed around it changes the format. This guide explains how a multilingual video meeting works when every participant speaks and listens in their own language, what to expect from the interpretation, and how to run one well.

What a multilingual meeting room actually does

In a conventional interpreted call, interpretation is an add-on channel. You pick "French" in a menu and hear one interpreter, who is hearing one floor language. Everyone else still has to speak that floor language.

In a multilingual meeting room, the room itself is the interpreter. Each participant sets two things when joining: the language they speak and the language they want to hear. From then on:

Every speaker is interpreted individually. When Adam speaks English, the room produces French for Marie and Lithuanian for Tomas at the same time. When Marie answers in French, Adam hears English.
Everyone hears the room in one language, their own. There is no floor language. A meeting with four languages is four parallel conversations that stay in sync.
Live captions run alongside the voice. Each participant sees the interpreted text in their listening language, which helps with names, numbers and anyone joining from a noisy place.

The interpretation lags the speaker by a few seconds, the same delay a human interpreter needs to hear a complete thought. Interpretwise shows a small "still hearing translation" marker on the speaker's tile until every listener has caught up, so the next person does not start talking over the interpretation. It is a small detail that turns out to matter more than any other for keeping the conversation civil.

Three ways to listen

Not everyone wants the same thing from interpretation. A native speaker with decent English may want the original voices and captions as a safety net. Someone joining from a train may want text only. A meeting room should let each person choose, and change their mind mid-meeting:

Voice interpretation. The AI voice in your language, with the original speaker kept softly underneath so you still hear tone, laughter and interruptions. You control the mix.
Subtitles only. Original audio at full volume, interpreted captions on screen. Common for people who understand the source language partly and want confirmation rather than replacement.
Original. No interpretation. Useful for the moments you speak the same language as the current speaker.

Switching between these takes one click and does not interrupt anyone else. Participants can also change the language they speak during the meeting, which matters for bilingual teams that drift between languages depending on who is talking.

Who it is for

The format fits any meeting where the people who matter do not share a strong common language:

Cross-border teams and subsidiaries. Weekly syncs where the German plant, the Polish office and the Lithuanian headquarters each speak their own language, instead of everyone speaking approximate English.
Customer and supplier conversations. Sales calls, technical reviews and negotiations where "we understood each other" needs to be true.
Town halls and all-hands. Leadership speaks once; every employee hears it in their language. See our guide on running a multilingual town hall.
Training and onboarding. A trainer delivers in one language and every trainee follows in theirs, with questions flowing back the same way. We cover the specifics in simultaneous interpretation for corporate training.
Hybrid panels and workshops. Remote panellists join a physical event and take part, not just watch. The venue joins the room as one participant through its sound desk, which we describe in the hybrid event interpretation guide.
Public bodies and associations. Committee meetings and member calls with participants from several countries, where a browser-based, GDPR-compliant platform is a requirement rather than a preference.

What to expect from AI interpretation quality

Set expectations honestly, because participants will notice the difference from a human interpreter within the first minute, and they will judge the whole meeting by whether the difference was explained.

Where AI does well. Factual, structured speech: status updates, presentations, technical discussion, Q&A with clear questions. Terminology is consistent, numbers and names come through, and the AI never tires in hour three. It also covers 80+ languages, including pairs such as Lithuanian to Japanese that would be nearly impossible to staff with human interpreters at short notice.

Where you need to manage it. Fast overlapping speech, heavy idiom, sarcasm and jokes. The AI interprets what was said, not what was meant. A human interpreter will smooth a rambling sentence; the AI will deliver the ramble faithfully. For high-stakes diplomacy, legal proceedings or medical consent, human interpretation still wins, and the best platforms let you add a human interpreter to specific languages while AI covers the rest.

Voice quality is a choice. A standard voice tier is clear and accurate but flatter; a premium tier sounds like a natural speaker with intonation and pauses. For internal working sessions the standard tier is fine. For customer-facing meetings, premium is worth the difference.

Speaker discipline matters more than the software. One person at a time, complete sentences, a beat of silence before the next speaker. The same rules that make a human-interpreted meeting work. Our speaker's guide for multilingual events applies word for word.

Setting up a meeting: what the host decides

A good meeting room asks the host for a handful of decisions up front and then gets out of the way.

Languages. Pick the set the meeting will run in. Participants choose from that set when they join. More languages cost more, so pick the ones people will actually use.
Admission. Anyone with the link, or invited participants only with a waiting room. For external meetings, use the waiting room.
Duration. The meeting is priced by the minute, so the scheduled duration is what gets reserved. If the meeting runs long, the host adds time from the company's balance without anyone leaving the room.
Listen-only attendees. For a town hall or a briefing, most people do not need a camera or a microphone. Listener seats let a larger audience follow in their language without turning the meeting into a crowd.
Recording. Optional per meeting. When enabled, each language is recorded as its own track, alongside a transcript and an AI-generated summary you can download as a PDF afterwards.

The full cost is shown before the host confirms, in the same screen, so there is never a surprise invoice. See the pricing page for the current rate card, or our conference interpretation cost breakdown for how meeting rooms compare with live event pricing.

Joining: what participants experience

Participants should not need an account, an app or a briefing. The join flow that works:

Open the link in any modern browser, on a laptop or a phone.
Pick a microphone and camera, and confirm the language you speak. The room suggests one based on your browser language or what you chose last time.
Pick what you want to hear: voice interpretation, subtitles only or the original audio.
Join. A short card shows who else is in the room and what each of them hears, so nobody wonders whether their French is actually reaching the Spanish colleague.

Hosts get the moderation tools you would expect: mute a participant or everyone, ask someone to unmute, lock the meeting, remove a participant, and set another participant's languages if they picked wrongly. Chat and reactions work regardless of language.

Reliability: audio first, always

A multilingual meeting fails differently from a normal one. If video stutters, people cope. If the interpretation drops for ten seconds, half the room has lost the thread and does not know it.

The right design treats interpreted audio and captions as the thing that must never stop, and video as the thing that can degrade. Under constrained bandwidth, video quality steps down first, then pauses entirely if needed, while audio and captions continue. Participants on a weak mobile connection still follow the meeting; they just see fewer faces. We go deeper on the technical side in the audio setup guide and the RSI platform security checklist.

What it costs

Meeting rooms are priced per hour and billed per started minute, with the first language included in the base rate and each additional language priced on top. At Interpretwise's standard rates, a two-language meeting costs less than a single human interpreter's minimum call-out, and a five-language, one-hour workshop costs roughly what a taxi to the venue would. The figures are on the pricing page with an estimator for your exact configuration.

Compare that with the traditional setup: two interpreters per language, minimum half-day rates, plus a technician and a booth or a remote interpreting platform on top. For the occasional high-stakes meeting, that investment is justified. For the Tuesday sync with three subsidiaries, it never was, which is why those meetings ran in broken English.

Hear it before you book it

The fastest way to judge AI interpretation is to listen to it on real speech. We publish recordings of live interpreted runs on our examples page: pick a language, press play and read the captions as they land, with the original audio available for comparison. Nothing there was corrected afterwards.

When you are ready to try it on your own meeting, contact us. We will set up a room with your languages, join the first call with you and stay on hand while your team learns the rhythm.

Frequently asked questions

Do participants need to install anything?

No. Meetings run in the browser on desktop and mobile. Guests join by link without creating an account.

How many languages can one meeting have?

As many as the host selects from the 80+ supported languages. Each participant hears the room in their own listening language; the number of languages in use does not change what any individual experiences.

Can I switch languages during the meeting?

Yes. Both the language you speak and the language you listen in can be changed mid-meeting. Listening changes apply immediately; speaking changes apply at the end of your current sentence so nothing is cut off.

Can I add a human interpreter for one language?

Yes. Human interpreters can cover specific languages from a browser-based console while AI covers the rest. This is the usual setup for a meeting where one language pair carries legal or commercial weight.

Is the meeting recorded?

Only if the host enables recording. Each language is then recorded as a separate track, and a transcript with an AI summary is produced afterwards. Participants are told the meeting is being recorded when they join.