Google showed off a range of new features for creators, developers, and everyday users at I/O 2025, beyond its headline announcements about search and AI models. The updates cover everything from AI-powered video and image generators to new XR platforms, agent tools, and learning options.
The company introduced Veo 3, a new AI video generator that can create videos with native audio tracks—including background sounds and dialogue. This sets Veo 3 apart from competitors like OpenAI's Sora, which doesn't support audio. Veo 3 is currently available to US users on the Ultra subscription tier and to enterprise customers through Vertex AI.
Alongside Veo 3, Google also launched Imagen 4, an upgraded image generator with improved detail, better typography, and support for multiple formats. Both Veo 3 and Imagen 4 are already integrated into several Google products, including the Gemini app, Workspace, Whisk, and Flow—a new AI filmmaking tool that can generate complex clips, scenes, and entire films. Flow brings together features like camera and perspective control, asset management, and the ability to mix user-generated or AI-generated content.
Google also opened up Lyria 2, its AI model for music composition, to musicians and producers via YouTube Shorts and Vertex AI.
AI Ultra and Pro: New subscriptions for power users
Google is launching a new high-end subscription, AI Ultra, for $249.99 per month. The plan offers top-tier access to all AI models, deep research tools, video, and image generation via Veo 3 and Imagen 4, agent features (Project Mariner), YouTube Premium, and 30 TB of storage.
The Gemini app is getting Gemini Live, a feature that enables real-time help through camera and screen sharing—useful for troubleshooting or event planning. Gemini is also being integrated into Chrome, letting users analyze and summarize website content with AI.
For Google Meet, a new AI-powered real-time translation feature is rolling out, which mimics the speaker's tone and expression. This will initially be available in English and Spanish for Pro and Ultra subscribers, with more languages to follow.
Gmail's personalized smart replies now use data from users' emails and Drive to generate responses in the user's personal style. The feature is opt-in and requires explicit user consent.
Developer tools: Gemini API, new agents, and open models
Google unveiled a host of new tools for developers. The Gemini API now supports advanced features like computer control (Project Mariner), native audio output, asynchronous function calling, and the Model Context Protocol (MCP) for integrating external tools. New models such as Gemma 3n (lightweight, multimodal, designed for mobile), MedGemma (for medical applications), and SignGemma (for sign language) expand the lineup.
On the coding side, new products like Gemini Code Assist, the asynchronous coding agent Jules, and the UI generator Stitch aim to automate and speed up the development process.
Android XR and SynthID
Google is rolling out the SynthID Detector, a portal that can identify AI-generated content—including images, text, audio, and video—using invisible watermarks. The technology is meant to help prevent misinformation and increase transparency across media. Google says more than 10 billion items have already been marked with SynthID.
The company also introduced Android XR, a new platform for augmented and mixed reality devices. With Gemini integration, XR glasses and headsets will get assistive features like live translation, navigation, messaging, and environment analysis from the user's perspective. Google is already partnering with eyewear brands like Gentle Monster and Warby Parker, and a developer SDK is planned for later this year.
