Google Meet's real-time voice translation jumped from six languages to 63 this week, powered by a new Gemini 3.5 Live Translate model that automatically detects which language someone is speaking instead of locking a call into one fixed pair. This expansion brings Google's Gemini 3.5 Live Translate model to Meet's web and mobile apps, and it removes the old meeting-wide language-pair limit so each participant can pick their own translation language. Most people will still need an eligible paid plan to use it.
Until this week, mobile translation only worked between English and five European languages, and only one pair could be active in a meeting at a time, Android Police reported. A French-speaking colleague joining an English-Spanish call simply wouldn't get translated audio under that system. That restriction is gone now, but using the feature still requires an updated Meet app and, for most people, an active Google AI or Workspace subscription.
How Google Meet real-time voice translation works now
Meet no longer requires anyone to preset a language pair before a call starts. The system listens to whoever is speaking, identifies the language automatically, and translates it into each listener's chosen language in near real time, even when several languages are in play during the same meeting or a speaker switches mid-call.
Participants must also consent to having their speech translated for others. If someone declines, other attendees won't hear translated audio from that speaker. Each participant turns live audio translation on individually through a dedicated control in the meeting toolbar, picks a language to hear, and can mark other languages they already understand so Meet skips translating those, the outlet notes. Captions update alongside the translated audio to keep reading and listening in sync.
The update also runs on different plumbing than what mobile users got six months ago. Older systems transcribed speech into text, translated that text, then converted it back into synthetic speech, a chain Google's engineers said produced 10 to 20 seconds of latency, far too slow for real conversation, in a 2025 engineering post.
That post described an end-to-end speech-to-speech approach, built with Google DeepMind, that could cut the delay to what the team called a two-to-three-second sweet spot for natural back-and-forth. That figure describes the earlier, narrower version of the feature; nothing in the available reporting confirms the 63-language expansion has been independently measured at the same speed.
How to use Google Meet voice translation on mobile
The 63-language version works on the Android and iOS Meet apps as well as the browser, with listening support also reaching Meet conference-room hardware. Since this is a fresh rollout tied to a specific app build, keeping Meet updated to its latest version is the safest first move before hunting for the new toolbar control.
Using it comes down to a short sequence, based on what's been reported so far:
Update the Google Meet app.
Join a meeting where you or the organizer holds an eligible plan.
Open the dedicated translation control in the meeting toolbar.
Choose the language to hear.
Optionally mark languages already understood so they're skipped.
Turn translation on; it doesn't activate automatically.
Google began rolling out the expansion to Rapid Release domains on October 22, with full visibility expected within one to three days. Scheduled Release domains begin receiving it November 2, with rollout potentially taking up to 15 days.
When Google first brought speech translation to mobile six months ago, it set the feature on by default for eligible Workspace editions, with admins able to toggle it off at the organizational-unit level, a Google Meet Help Community post explained. Google confirms that speech translation remains enabled by default for eligible Workspace editions when the Gemini for Meet admin setting is enabled. Admins can manage access by organizational unit or group, while individual participants must turn translation on during a meeting.
Which plans actually qualify
This remains a paid-tier feature. A standard free personal Google account doesn't unlock it, Android Police reported five months ago, and nothing in current reporting suggests that's changed. Eligible plans include:
Consumers: Google AI Pro and Google AI Ultra
Businesses: Business Standard or Plus, Enterprise Standard or Plus, Frontline Plus
Add-ons: AI Expanded Access and AI Ultra Access
Education: Google AI Pro for Education
Subscription eligibility doesn't guarantee access. Google also restricts speech translation by region and age, including for users under 18. Workspace administrators can disable the feature for their organizations.
A meeting organized by someone on an eligible plan opens translation to every participant, including guests from outside domains. In a one-on-one call, either person holding an eligible plan covers both sides, the outlet notes.
Time limits still scale with tier. Base Workspace and Google AI Pro accounts get 90 minutes of translation per meeting, while AI Expanded Access, AI Ultra Access, and Google AI Ultra accounts get up to eight hours per meeting.
Where the feature still falls short
Translation doesn't carry over to livestreams, meeting recordings, client-side encrypted meetings, dial-in calls, third-party interoperability calls, or any meeting with adaptive audio turned on, Android Authority notes.
Conference rooms running Meet hardware get a partial version of this. People in the room can hear translated speech from remote participants, but their own speech isn't translated for anyone listening elsewhere, per the outlet. Hybrid meetings mixing in-room and remote attendees should expect that one-way gap.
The most detailed account of where this kind of system struggles still comes from Google's engineers discussing the earlier, narrower version of the feature. They said accents, background noise, network conditions, and structurally different languages such as German created real challenges, and that the system tended to translate idioms literally, sometimes with odd results, in the 2025 engineering post.
Android Police also noted that Google still labeled the feature a beta, meaning behavior could keep shifting. Nothing in the available reporting confirms whether that beta label still applies to the 63-language expansion, or whether independent benchmarks exist for it yet.
What to do now
Update Google Meet first, then confirm whether you, your meeting's organizer, or your one-on-one partner actually holds one of the eligible Google AI or Workspace plans listed above. If the translation control isn't showing up, check your rollout schedule, account eligibility, regional availability, and Workspace admin settings before assuming the feature simply hasn't reached your device.
Given Google's own account of accent, noise, network, and idiom-related limitations in the earlier version of this feature, test the toggle on a routine call before leaning on it for anything where a mistranslation would actually matter.
Comments
Be the first, drop a comment!