In a groundbreaking move announced at Google I/O 2025, Google is transforming global communication by introducing real-time speech translation to Google Meet. This powerful new feature uses an advanced large language audio model from Google DeepMind, designed to make conversations across different languages feel natural and effortless.

Breaking Language Barriers Like Never Before

The feature works by translating spoken words into the listener’s preferred language in real time, while keeping the original voice, tone, and expression intact. This means conversations will still sound human and emotional, not robotic or flat.

According to Google, this innovation has many practical uses. Imagine English-speaking grandchildren having a smooth chat with their Spanish-speaking grandparents, or international teams in companies communicating instantly, no matter what language they speak. The possibilities are endless. What makes this feature truly game-changing is its ultra-low latency. Conversations happen almost instantly, making real-time group discussions across languages not only possible, but smooth and seamless, something Google says “hasn’t been possible until now.”

Even more impressive, when someone speaks, you’ll still hear their original voice faintly, with the translated speech overlaid on top, preserving the authenticity of the speaker.

Rolling Out Soon with Support for More Languages

This new translation tool will begin rolling out in beta to consumer AI subscribers on Tuesday. Initially, it will support English and Spanish, with Italian, German, and Portuguese joining in the next few weeks. Looking ahead, Google is also expanding this capability to the business world. Workspace customers will get early access for testing later this year, as the company continues to refine the feature for professional use.