Google I/O | AI-powered Beam, Google Meet translation, and Project Astra integration introduced

Google further added that it was bringing agentic capabilities to Chrome, Search and the Gemini app [File]
| Photo Credit: REUTERS

At its annual I/O event for developers and other technologists, Google launched a range of Generative AI-powered offerings, including a 2D-to-3D video communication platform called Beam, more natural speech translation in Google Meet, and Project Astra’s integration into Gemini Live.

Beam is a device being built in collaboration with HP and it will be available for early customers later this year. Beam uses a video model to turn 2D video streams into a realistic 3D experience, per Google, using cameras and AI to render subjects on a 3D lightfield display.

Meanwhile, Google AI Pro and Ultra subscribers in beta will get near real-time translations in Google Meet for the English and Spanish languages. More languages will roll out later. Workspace business customers will get the feature for early testing this year. The new AI translation feature is significant as it will aim to match the speaker’s voice, tone, and expressions to make the translated interaction feel more organic.

In addition to this, Google’s Project Astra AI assistant offering is being integrated into Gemini, so that Gemini Live users can access Project Astra’s camera and screen-sharing capabilities. Android users can try out the feature already while iOS users are in the process of receiving it.

Google further added that it was bringing agentic capabilities to Chrome, Search, and the Gemini app for more personalised and autonomous search experiences online.

Google CEO Sundar Pichai also highlighted Google’s pace of shipping new releases, rather than waiting for special occasions or events.

“I’m particularly excited about the rapid model progress. Elo scores, a measure of progress, are up more than 300 points since our first-generation Gemini Pro model. Today, Gemini 2.5 Pro sweeps the LMArena leaderboard in all categories,” said Pichai in his speech.

“Model progress is enabled by our world-leading infrastructure. Our seventh-generation TPU, Ironwood, is the first designed specifically to power thinking and inferential AI workloads at scale. It delivers 10 times the performance over the previous generation, and packs an incredible 42.5 exaflops compute per pod — just amazing,” he added.