Google Meet Introduces Real-Time Speech Translation Powered by DeepMind

At Google I/O 2025, Google announced a groundbreaking enhancement to Google Meet: real-time speech translation powered by a large language audio model developed by Google DeepMind. This feature enables users to engage in natural, fluid conversations across different languages without losing the nuances of voice, tone, or expression.

Unlike traditional translation tools, this technology overlays the translated speech on the speaker’s original voice, allowing listeners to faintly hear the original audio while receiving the translation in their preferred language. This approach preserves the emotional and tonal context of conversations, making interactions more authentic and engaging.

The feature is designed for diverse use cases, including personal connections like English-speaking grandchildren communicating with Spanish-speaking grandparents, as well as professional environments where global teams need seamless real-time communication. The low latency of the translation supports multi-person conversations, a capability previously unattainable with real-time speech translation.

Initially, the feature will roll out in beta to consumer AI subscribers starting Tuesday, supporting English and Spanish languages. Additional languages such as Italian, German, and Portuguese will be added in the coming weeks. Google is also preparing to extend this capability to Workspace customers, enabling businesses to benefit from real-time multilingual communication.

Broader Significance and Opportunities

This advancement marks a significant step forward in AI-driven communication tools. By integrating sophisticated audio language models, Google Meet is setting a new standard for inclusivity and efficiency in virtual meetings. Businesses operating across different regions can now foster deeper collaboration without language barriers, accelerating decision-making and innovation.

For developers and tech leaders, this opens opportunities to explore how real-time speech translation can be integrated into other communication platforms and services, enhancing accessibility and user experience globally.

Conclusion

Google Meet’s real-time speech translation feature powered by DeepMind’s AI represents a major innovation in global communication. By preserving the natural qualities of speech and enabling low-latency multilingual conversations, it promises to transform both personal and professional interactions worldwide.