Google Meet now translates speech live in 63 languages. The meetings that need it most are the ones it skips
Gemini speech translation in Meet now covers 63 languages with automatic detection. It is off in client-side encrypted meetings and recordings, so regulated teams need a policy, not a toggle.
Listen to this article · 5 min
AI-generated narration of the full article.
For multinational companies, the expensive part of a global meeting has rarely been the video. It is the language. A plant manager in Monterrey, a quality lead in Campinas and a regulatory specialist in Basel can all speak English, and still the meeting runs at the pace of whoever is least comfortable in it.
On October 22 Google expanded real-time speech translation in Meet to 63 languages, powered by Gemini 3.5 Live Translate. The change is not only the number. Previously, a meeting had one language pair, set for everyone. Now Meet detects the language being spoken and translates it into each listener’s chosen language, even when speakers switch languages mid-sentence, and each participant turns translation on for themselves. Translated captions follow the translated audio.
The details that decide whether you can use it
- Who gets it. Every participant, including guests from other companies, when the organizer has an eligible Workspace or Google AI plan.
- How long. Up to 90 minutes per meeting on eligible base Workspace and Google AI Pro plans, and up to 8 hours on the higher AI tiers.
- Rollout. Rapid Release domains from October 22, Scheduled Release from November 2, with up to 15 days for features to appear.
- Default settings. On by default for eligible editions when the Gemini for Meet setting is enabled; administrators can turn it off by organizational unit or group. For users, it starts off in every meeting.
- Privacy. Users choose whether their own speech is translated for others, and Google says it does not store voice data or train models on users’ voices.
Where it does not work
Google lists the scenarios where speech translation is unavailable: live streams, meeting recordings, dial-in and third-party interoperability calls, adaptive audio, and client-side encrypted meetings. In conference rooms with Meet hardware, people can hear translated speech from remote participants, but what is said in the room is not translated for others. Availability is also subject to regional and age restrictions.
The client-side encryption exclusion is the one to notice. Organizations turn on client-side encryption precisely for their most sensitive conversations: board discussions, legal strategy, regulated data. Those are the meetings where translation would be most valuable and where it is not available. That is a defensible engineering choice, because translating speech requires processing it, which client-side encryption is designed to prevent. It does mean the feature reaches everyday collaboration first and sensitive collaboration last.
Translation in the meeting is not translation in the record
Two practical consequences follow for companies in regulated industries.
The record stays in the original language. Translation does not work in recordings, so the official record of a multilingual meeting will be the original audio and whatever notes or transcripts the company keeps. If a decision was made by someone who understood it through live translation, the company should know that and, for consequential decisions, confirm it in writing in the language that person reads.
Accuracy is not guaranteed where it matters most. Live translation of technical, legal or medical language can be wrong in ways a listener cannot detect. For a project status call, that is acceptable. For a safety discussion, a negotiation or a regulatory interaction, it is a reason to use a human interpreter or to confirm the key points in writing. Microsoft’s support organization, which uses AI translation to connect customers in more than 60 languages to scarce experts, treats it as a guarded system with requirements revisited as the technology changes, not as a toggle.
What administrators should decide now
- Choose the default by group. It is on by default for eligible editions. Decide whether that is right for legal, HR or clinical teams, or whether they should opt in.
- Tell people what translation is for: understanding a colleague, not creating a record or agreeing to terms.
- Write a rule for consequential meetings: decisions taken with live translation are confirmed in writing, in each participant’s language.
- Plan for the encrypted-meeting gap, so that teams that need both confidentiality and multiple languages know their option is an interpreter.
Live translation is one of the clearest everyday wins for AI in multinational companies, and in Latin America and Europe it will be used daily within weeks. The work is deciding where a translated meeting is good enough, and where it is not a substitute for words both sides can read.