Google has begun rolling out new real-time AI video features for Gemini, its AI assistant, allowing it to interpret and respond to content on users' smartphone screens and camera feeds. This development is part of Google's Project Astra, which was first demonstrated nearly a year ago. The features are currently available to Gemini Advanced Subscribers of the Google One AI Premium plan, with initial rollout targeting Pixel devices, some Samsung phones, and other Android devices including Xiaomi phones. The new capabilities enable Gemini to provide immediate, context-aware assistance based on visual input. Users can point their phone cameras at objects or scenes, and Gemini will analyze the visual data to offer relevant information or answer questions in real-time. This marks an advancement in integrating AI into everyday tasks, enhancing user interaction with technology. In addition to screen and camera analysis, Google is preparing to unveil a new model for Gemini this week, which will include a toolbox for 'Agents.' These agentic use cases are expected to further enhance Gemini's capabilities, potentially integrating with existing features like Canvas and Deep Research.