Google has launched a new capability for its Gemini AI assistant, enabling it to generate personalized images by accessing a user's Google Photos library. This update is part of Google's broader Personal Intelligence initiative, which allows Gemini to utilize data from connected apps to deliver responses tailored to individual contexts. The feature integrates the system with the Nano Banana 2 image model, allowing it to create visuals that reflect a user's specific tastes and lifestyle. Users can now issue prompts such as "Design my dream house" or "Create a picture of my desert island essentials." In response, Gemini analyzes images and labels within the user's Google Photos to understand their environment and preferences. The system identifies people such as the user, friends, and family through these labels before the Nano Banana 2 model generates the requested image. This process aims to produce results that feel uniquely aligned with the user's personal history and style. Regarding privacy, Google clarified its approach to data handling. While the company does not directly train its underlying AI models on the contents of a user's private Google Photos library, it does use limited information for system improvement. This includes specific prompts entered into Gemini and the model's subsequent responses. Google emphasizes that the image generation relies on real-time context retrieval rather than permanent model training on private visual data. The rollout of this feature is scheduled to begin over the next few days. Initially, it will be available to eligible subscribers of AI Plus, Pro, and Ultra plans within the United States. Following this initial phase, Google plans to expand access to Gemini on Chrome desktops and eventually to a broader range of users. This expansion underscores Google's strategy to deepen the integration of its AI services with existing user ecosystems to provide more relevant and personalized experiences. The move represents a significant step in making AI-generated content more dynamic and context-aware for everyday users.