Google Meet Introduces Real-Time Speech Translation

Google announced at I/O 2025 that Google Meet is gaining a powerful new feature: real-time speech translation. This innovation, powered by a large language audio model from Google DeepMind, aims to enable seamless, natural conversations between people speaking different languages.

The new feature translates spoken words into the listener's preferred language instantly, while remarkably preserving the original voice, tone, and expression. Google highlights diverse use cases, from English-speaking grandchildren chatting with their Spanish-speaking grandparents to global colleagues in multi-regional companies connecting effortlessly in real time.

Google emphasizes the very low latency of this speech translation, which allows for multiple participants to converse simultaneously—a capability the company claims hasn't been possible until now. Users will faintly hear the original speaker's voice overlaid with the translated speech.

Real-time speech translation will begin rolling out in beta to consumer AI subscribers starting this Tuesday. Initially, it will support English and Spanish, with additional languages like Italian, German, and Portuguese coming soon. Google also plans to bring this feature to businesses, with early testing for Workspace customers slated for later this year.