How to Translate a Zoom or Teams Call Live

Start with the platform you are already on. Zoom, Microsoft Teams and Google Meet all translate live, and for most people the native feature is the right answer because it needs no extra software and no second device. The catch is the gating, which nobody spells out: translated captions and translated speech are different features on different paid tiers, and on two of the three platforms somebody other than you has to switch something on before you can use them. If you are a guest on a call inside another company's account, none of that is available to you, and the fix is a separate audio channel running beside the meeting window rather than inside it.

I sell translation earbuds and a browser translation app, so treat me as an interested party. That is exactly why the first half of this page is about the platforms rather than about my product. If the native path works for your call, use the native path. It is cheaper, it is in the window you are already looking at, and it does not ask the other person to change anything.

Everything below about Zoom, Teams and Meet was read on each vendor's own current support pages in September 2026. Plan names and tiers in this category change every few months, so check the linked page before you buy anything on the strength of a blog post, including this one.

What can Zoom, Teams and Meet translate natively?

All three, and more than most people realise. The differences are in what comes out, and who is allowed to turn it on.

Platform	What is native	Who has to turn it on	Captions or spoken audio	If you do not control the account
Zoom	Translated captions, plus a separate voice translator that speaks a synthetic translation	Account owner or admin enables translated captions for the account or a group, or a user enables it on their own account unless an admin has locked it. Participants then pick their own language mid-meeting	Both, on two different paid paths	Nothing you can enable from the guest side. Run a second audio channel
Microsoft Teams	Live captions for everyone, live translated captions, and the Interpreter agent for speech to speech	The organiser sets the available translation languages in meeting options before the event, and the licence that pays for it sits with the organiser. Interpreter is switched on per person by each attendee	Both. Captions are text, Interpreter speaks over the original voice	Ask the organiser, or run a second audio channel
Google Meet	Translated captions	Each participant turns them on in their own settings, but the feature is gated by the Workspace edition on the account	Text captions only. No spoken output	Run a second audio channel

Zoom's support documentation describes translated captions as letting users "have the speech in a meeting or webinar automatically translated in real-time to captions in another language", and lists the required plans as Business Plus, Enterprise Essentials, Enterprise Plus or Enterprise Premier, or a Translated Captions add-on on top of a lower plan. The details, including which language pairs are excluded, are on Zoom's page for enabling and configuring translated captions. Two constraints there are worth knowing before you plan a call around it: dialects of the same language cannot be translated into each other, and a handful of languages are supported as a translation target only, not as a language somebody can speak into.

Zoom also has a genuine speech to speech feature now. Its article on using the voice translator for meetings describes automated text to speech output in multiple languages, driven from the More menu inside a meeting where you set your speaking language and the language you want out. The prerequisites listed there include a paid translation add-on and a recent version of the Zoom desktop app, and the feature has to be enabled before it appears. So it exists, it is real, and it is not simply on.

Teams splits the same problem three ways. Ordinary live captions are the baseline, turned on from the meeting controls under Language and speech. Translated captions are the paid layer, and Microsoft's page on using live captions in Teams meetings states that they are "available as part of Teams Premium and Microsoft Copilot, an add-on license". The important half of that sentence is what follows it: if the organiser holds the licence, every participant can use translated captions, licensed or not. If only you hold it, only you get it.

Spoken translation in Teams is the Interpreter agent, and Microsoft's page on Interpreter in Teams meetings and calls documents it as real time speech to speech, switched on by each person for themselves with "Turn on for me", included with a Microsoft Copilot licence on a monthly hours allowance. It can also speak in a simulation of your own voice if an admin policy allows it. Two limits from that page matter for planning: Microsoft says it "isn't optimized for meetings with rapid exchanges, interruptions, or overlapping dialogue", and meeting recordings capture only the original audio, not the interpretation.

Google Meet is the simplest to describe and the most limited. Google's Help documentation on using translated captions in Google Meet lists the editions that include it, Business Standard and Business Plus through the Enterprise tiers among them, and the languages available on both sides of the translation. Each participant switches it on for themselves under Settings, then Captions, then Translated captions. What comes out is text on screen. There is no spoken translation in the product.

Captions or spoken audio, which do you need?

This is the fork that decides everything else, and it gets skipped constantly.

Captions are text. You read them. They are excellent when the call is one presenter talking steadily and you need to follow along, when you want a scrollback you can re-read after a term you did not catch, and when you are in a room where you cannot play audio. They are also the cheaper tier almost everywhere, and on Google Meet they are the only tier.

Captions are poor when you need to talk back quickly, because reading and speaking compete for the same attention. They are poor when you are also sharing a screen, since the captions and the deck fight for the same rectangle. And they are poor on a phone, where the caption strip eats the video and the text is small.

Spoken translation is the opposite. You keep your eyes on the person, the document or the whiteboard, and the translation arrives in your ear. That is what you want for a back and forth negotiation, an interview, or a site walkthrough on video. What you pay for it is latency and overlap. The translated voice starts after the original speaker has already begun, so you are always hearing the previous sentence while the next one is being said. Microsoft says this plainly in its own limitation note, and it applies to every speech to speech system including mine.

A practical rule that has held up for me: if the call is mostly one person explaining, take captions. If the call is a conversation with turns in it, take audio.

Who has to turn it on, you or the host?

Three different answers, which is why so many people conclude their platform "doesn't have it" when it does.

On Google Meet, you turn it on yourself, in your own settings, mid-call. But the ability to do that comes from the Workspace edition of the account you are signed in with. Personal Gmail accounts and lower business tiers do not get it, and no amount of clicking finds a switch that is not there.

On Zoom, translated captions have to exist at the account level first. An owner or admin enables them for the account or for a group, or an individual user enables them on their own account if an admin has not locked the setting. Once they are live, participants pick their own caption language during the meeting without asking the host for anything.

On Teams, the organiser is the gatekeeper for translated captions. The languages available in a meeting are chosen in that meeting's options before the event, and the licence that pays for the feature is the organiser's. Interpreter works the other way around: it is a per person switch, and it depends on your own Copilot licence, not the organiser's.

So the honest advice for anyone inside an organisation is not "buy something". It is: ask your IT administrator whether the account already has the tier, then ask the organiser to set the languages in the invite. In a lot of companies the licence is already paid for and nobody has ever opened the setting.

What do you do on someone else's call?

Here is the case none of the vendor pages answer, because from their side it does not exist. You have been sent a link. The meeting is in another company's tenant. You are a guest, you cannot see the account settings, the host has not enabled translated captions, and the call is in ninety minutes.

Your realistic options are three.

Ask. Genuinely the best one, and the least used. A one line message to the organiser, saying you would follow better with translated captions on and asking whether their plan includes them, costs nothing and often works. It also tells them something useful about their own meeting.

Change your seat in the meeting. If you can dial in on a second device, or join twice, you have more room to work with than you think. That is the basis of the next section.

Run a separate audio channel beside the meeting. Translation happens outside the meeting window, on a device or a tab you do control, and nothing about the host's account has to change. This is where a browser tool or a pair of translation earbuds earns its place, and it is the only one of the three that works when the answer to the first option is no.

I will be direct about the trade. A separate channel is never as tidy as a native feature. The meeting has no idea it is happening, so nothing appears in the recording, the transcript or the chat, and there is no shared caption everyone reads together. If you have a recurring internal meeting with the same people every week, do not build a workaround around it. Buy the add-on or the licence, get it enabled properly, and the problem is gone permanently. The separate channel is for the one off call, the guest seat and the client tenant you will never have admin rights in.

How do you run a second audio channel cleanly?

The setup that actually works is two devices, not one, and understanding why saves you an hour of confusion.

A translation tool running in a browser tab hears what your microphone hears. It cannot reach inside Zoom or Teams and tap the call's audio stream, because browsers do not hand out other applications' audio, and every serious meeting client runs echo cancellation that actively removes your own speakers' output from anything your microphone picks up. That is a feature, not a bug, and it is what stops calls from howling. But it means the tidy sounding plan, one laptop running the call and a tab translating it, quietly fails on the inbound half.

What works:

Outbound, one device. Speak into the translation tab, and let it speak the translation out loud into the meeting, or read the translated text aloud yourself. Your voice reaches the other side in their language. This half needs nothing special.

Inbound, two devices. Join the meeting on your laptop with the call audio coming out of the laptop speakers, and run the translation on your phone next to it. The phone's microphone hears the room, the room contains the other person's voice, and the translation arrives in the earbud in your ear. It looks slightly ridiculous and it works reliably, which is a fair description of most useful workarounds.

Or hand over an earbud. If the person you need to understand is in the room with you and the call is the third party, one bud each is faster than any software route.

AI Translation Live is built for exactly this shape. It runs in a browser tab next to the meeting window, so on a managed work laptop where you cannot install anything, there is nothing to install. It works with any headphones you already own, including the ones you are wearing for the call. The audio goes from your browser straight to the translation model over an encrypted connection rather than through our servers, and no recording is kept, which is a sentence worth being able to say to a compliance team before you use anything on a client call. The kit we sell is a set of earbuds plus that app, and the earbuds are ordinary hardware whose job is to put a microphone next to a mouth instead of on a table.

If your meeting need is mostly this, and you are indifferent to hardware, the app on its own does the job. The comparison in browser translation versus an installed app covers where each side loses, including the case where a browser is the wrong architecture entirely.

What breaks when several people speak?

Every live translation system, mine included, degrades the moment two people talk over each other. It is worth knowing exactly how, because it changes how you run the call rather than which product you buy.

Speech translation works on completed thoughts. The system waits for enough of a phrase to know what it means, translates it, then speaks or prints it. When two voices overlap, the recogniser gets one signal containing both, and it either picks the louder one and loses the other, or produces a hybrid of the two that is worse than either. Microsoft's own documentation admits this about Interpreter, and it is the same physics for everyone.

Three consequences for a real meeting.

The lag is cumulative, not fixed. A speaker who does not pause runs further and further ahead of the translation, so a long uninterrupted monologue is harder to follow than a series of short turns, even though it sounds easier.

Cross talk destroys accuracy faster than accent or background noise does. If you are chairing a multilingual call, the single highest value thing you can do is enforce one speaker at a time and a beat of silence between turns. That is worth more than any setting.

Recordings do not save you. Teams captures only the original audio in the recording, not the interpretation, and a caption transcript is only as good as the live recognition was. If the call matters, do not plan to reconstruct it afterwards.

None of this is a reason to skip live translation. It is a reason to expect roughly the quality of a competent but tired interpreter rather than a perfect transcript, and to run the meeting in a way that plays to it. For anything with legal or medical consequence, use a human interpreter. I would rather say that here than have you find out during the call.

If your real problem turns out to be recurring meetings rather than a one off, translation earbuds for business meetings goes through the room setup and the microphone placement side of it, and how translation earbuds actually work explains why capture quality decides more of the outcome than the translation model does.

Frequently asked questions

Can Zoom translate a meeting in real time?

Yes, in two separate ways. Translated captions turn speech into text in another language on screen, and Zoom's documentation as of September 2026 lists them as part of Business Plus, Enterprise Essentials, Enterprise Plus and Enterprise Premier plans, or as a paid add-on to a lower plan. Separately, the voice translator produces spoken translated audio, and its prerequisites include a translation add-on and a current version of the Zoom desktop app. Neither is on by default. Translated captions have to be enabled at account or user level first, after which participants choose their own language during the meeting.

Does Microsoft Teams translate speech live?

Yes. Teams has live translated captions, which are text, and the Interpreter agent, which is real time speech to speech and plays a translated voice over the original speaker. Microsoft's support pages state that translated captions come with Teams Premium or the Microsoft Copilot add-on licence, and that if the organiser holds the licence all participants can use them. Interpreter runs on a Copilot licence with a monthly hours allowance and is switched on by each person individually. Microsoft notes that Interpreter is not optimised for rapid exchanges or overlapping dialogue.

How do I translate a call I didn't organise?

You cannot enable the host's features from a guest seat, so you have two workable routes. Ask the organiser whether their plan includes translated captions and whether they will set the languages in the meeting options, which takes them under a minute. If that is not possible, run translation outside the meeting on a channel you control: join the call on your laptop with audio playing out of its speakers, run a translation tool on a phone beside it so its microphone hears the room, and listen through an earbud. Meeting clients strip their own output from your microphone, which is why a single device cannot translate the incoming half of a call on its own. You can try the browser side of that setup on the AI Earbuds home page before you commit to anything.

Do meeting translations get saved?

Usually less than you expect. Microsoft states that Teams meeting recordings capture the original meeting audio only and do not include the interpretation audio. Caption transcripts, where the platform saves them, preserve only what the live recogniser produced at the time, mistakes included, and a separate channel running outside the meeting leaves no trace in the meeting record at all. In our own app, audio streams from the browser straight to the translation model and no recording is kept, which is the right default for privacy and the wrong one if you needed a record. If the call has to be minuted accurately, assign a person to take notes.