StreetViewAI: Google’s Multimodal AI Brings Conversational Street View Navigation to the Visually Impaired

Google Research and DeepMind introduced StreetViewAI, a multimodal, conversational system built on Gemini Flash 2.0 that makes Google Street View accessible to visually impaired users via three components: an AI Describer, AI Chat Agent, and AI Tour Guide. In a study with 11 visually impaired participants, the AI achieved 86.3% accuracy with only 3.9% incorrect answers, and users favored conversational interactions (917 chat queries vs. 136 descriptions). The system provides rich spatial, accessibility, and contextual information—such as verifying landmarks or accessibility features—and aims to enhance inclusivity with potential applications in education, tourism, and smart city navigation.