Google's Project Astra can control phones, make calls, and see through lens

At Google I/O 2025, Project Astra showed off next-gen AI skills. In the demo, the AI-powered tool was able to read screens to make calls and remember context shared earlier

The keynote session of Google I/O 2025 was headlined by how artificial intelligence (AI) powered Google tools have evolved in a year to effectively assist users across the globe. During the keynote session, Google shared insights into its Project Astra, which was announced at the Google I/O 2024 event. The US technology company shared how far the multimodal, all-seeing AI bot has come.

What is Project Astra?

Project Astra, created by Google DeepMind, is a multimodal AI system built to interact with users across various formats—text, voice, visuals, and video. Its goal is to function as an all-purpose AI assistant, providing real-time answers by interpreting both visual context and user input. Blending real-world observations with online data, Astra is designed to create more fluid, intuitive exchanges that mirror human communication and understanding.

 

In other words, the technology powering Gemini Live’s camera and screen sharing is called Project Astra. Google, while sharing the long-term goal with Project Astra, in a blog post wrote: “Our ultimate vision is to transform the Gemini app into a universal AI assistant that will perform everyday tasks for us, take care of our mundane admin, surface delightful new recommendations, making us more productive and enriching our.”

Project Astra at Google I/O 2025

Google showcased the latest features of Project Astra on a Pixel 9 Pro, demonstrating its ability to assist users in real-time through a range of tasks. In one scenario, a user was fixing their bicycle and asked Astra to find a repair manual. Astra promptly located a PDF, opened it in a floating bubble interface, and navigated through the file by scrolling to the relevant section, visually highlighting its interaction on the screen.

 

The demo continued with Astra launching the YouTube app, Browse search results, and playing a selected video. It also accessed the user’s Gmail to retrieve specific details from a past conversation. When shown a scene with multiple containers holding parts, Astra identified and marked the correct one, echoing earlier visual recognition demos from Google Lens.

 

Another standout feature was Astra’s ability to make phone calls for the user. It contacted a bike shop to inquire about a part, and even followed up to confirm the user’s interest before completing the request.

Astra also demonstrated its contextual awareness. When the user was interrupted mid-response by someone else, Astra paused automatically and resumed the conversation seamlessly when prompted. Later, when asked for a bike basket suitable for a pet, Astra remembered the dog’s name from earlier interactions, highlighting its long-term memory capabilities.

 

Google refers to this suite of functions as “Action Intelligence,” powered by Gemini. It combines features like content retrieval, interface navigation, agent-based interaction, voice calling, contextual awareness, and personalised recommendations.