Google I/O 2025 Developer Keynote: Gemini, Android, and AI Innovations

AI Innovations with Gemini

• 💡 Gemini 2.5 Flash now features a native audio model that better ignores stray sounds and supports 24 languages.
• 🧠 Project Astra capabilities, making AI feel natural, are now available via a live API.
• 🤖 A new AI agent in Android Studio will assist with version updates by fixing build problems using Gemini.
• 📸 Gemini Nano unlocks new multimodal capabilities, allowing interaction with audio and image input, demonstrated by instantly locating a seat section from a ticket photo.
• 🚀 Gemma 3N is announced, a model capable of running on as little as 2 GB of RAM, offering faster performance on mobile hardware and adding audio understanding.
• 💬 S-Gemini is a new model family trained to translate sign language to spoken language text, with a focus on American Sign Language.

Adaptive and XR Development

• 🛠️ New features in the compose adaptive layouts library, like pane expansion, simplify adaptive building.
• 👓 Development is expanding to Android XR, the extended reality platform built with Samsung, powering headsets like Project Muhan.

Development Tools and Integrations

• 🎨 Figma designs can be brought to life in Firebase Studio with help from Builder.io, generating isolated components for web pages.
• 📈 The Figma export generates a fully working product detail page from design to implementation.