Google I/O 2025: Google Meet can now translate speech in real-time using AI

At Google I/O 2025, Google unveiled a powerful new feature for Google Meet: real-time speech translation. Powered by a large language audio model developed by Google DeepMind, the new Gemini-based capability enables fluid multilingual conversations by translating spoken words into a listener’s preferred language, while preserving the speaker’s voice, tone, and emotional inflection.

Unlike Meet’s existing real-time caption translations, this new feature goes a step further by performing direct audio-to-audio transformations. That means users will hear their conversation partner's voice, not a robotic substitute, dubbed in another language with natural-sounding vocal cues. In a live demo, an English speaker and a Spanish speaker were able to seamlessly converse, with each hearing a translated version of the other’s speech overlaid in their native tongue. A faint version of the original voice plays in the background, allowing for added clarity and context.

The system is built on Google DeepMind’s AudioLM, trained specifically for high-fidelity speech translation. The result is low-latency dubbing that feels like having a live interpreter in the call, although Google does note there will be slight delays due to processing.

To use the feature, users can open the “Speech translation with Gemini” panel in Meet’s web interface, where they can specify both the language they are speaking and the one they wish to hear. Once enabled, the interface will display a glowing Gemini overlay and a translation indicator in the top corner of the screen.

The feature is now rolling out in beta to users on Google’s AI Pro and the newly launched AI Ultra plans, and the latter is priced at $249.99 per month. For speech translation to work, only one participant in a call needs to be subscribed, making it more accessible in group settings.

Initially, the feature supports English and Spanish, with support for other languages expected in the coming weeks. Google also confirmed that early testing for business users will begin later this year for Workspace customers.

With Microsoft Teams also experimenting with AI-powered translations, Google’s move signals a growing trend toward real-time, AI-assisted communication across language barriers.