Microsoft Teams live translation (2026): how it works, what it costs, and where it stops
If you searched "Microsoft Teams live translation" or "can Teams translate in real time," the honest short answer is: yes, Teams can translate a live meeting — three different ways — and which one you get depends on whose license the organizer has and what kind of meeting you booked. Microsoft has shipped more live-translation machinery than any other conferencing vendor, including a genuinely ambitious AI interpreter. It has also fenced each piece differently, and the fences are where buying decisions actually happen. This post explains how each works, how to turn it on, and exactly where the ceiling is.
This is the platform how-to companion to our foundational guide, Real-time meeting translation: how it works, and how to evaluate one. For the Google Meet and Zoom versions of this post, see Meet and Zoom. New to interpretation as a category? Start with Simultaneous interpretation: booth, RSI, or AI.
The three features, and why they're not the same thing
1. Live translated captions (text)
The established feature. Teams transcribes the meeting live and translates the caption stream into each participant's chosen language — every attendee picks their own caption language independently. Microsoft's support page lists 31 translation languages for meetings (its own licensing page says 40, and the town-hall page says "over 50" — we'd cite the 31-language list it actually enumerates). Works in meetings, webinars, and town halls — though in town halls the organizer pre-selects a pool of six languages, ten with Premium.
The gate: the organizer needs a Teams Premium or Microsoft 365 Copilot license for attendees to get translated captions. Plain English captions are free; translation is the paid layer. And captions aren't saved — when the meeting ends, the translation evaporates.
2. Interpreter agent (audio, AI)
The Ignite 2024 headliner, now generally available: an AI agent that translates spoken audio into spoken audio, per participant — you pick the language you want to listen in, and Interpreter can even simulate the speaker's own voice in the translation. Calls got it in January 2026; a turn-based consecutive mode for two-language meetings rolled out in May 2026. Of the big three platforms, this is architecturally the closest thing to per-listener simultaneous interpretation.
Which is exactly why its fences matter — they're in the limits section, and they're decisive.
3. Language interpretation (humans, channels)
The classic: the organizer pre-configures up to 16 language pairs and invites human interpreters — people you source, brief, and pay — who each get an audio channel. Attendees pick a channel and balance original vs. interpreter volume with a slider. No premium license documented, but the production burden is yours, and the constraints are real: set up before the meeting starts, interpreters must join from the desktop app, no web support, no breakout rooms, no end-to-end-encrypted meetings.