Google I/O 2025 key highlights: From Gemini upgrades to AI-powered shopping, here are 15 big announcements from the keynote

Virtual try-on for clothes in Search
Users can now upload a photo to try on clothes virtually. The system simulates how fabrics stretch and fold on different body types.

Google AI Pro replaces AI Premium
Google’s AI subscription tier is now called Google AI Pro, retaining its $19.99/month price. It offers higher usage limits and early access to new tools.

Google AI Ultra tier launches at Rs 24,500 per month
Aimed at power users and early adopters, Google AI Ultra provides top-tier access to unreleased features, including Gemini 2.5 Deep Think and Project Mariner.

Gemini 2.5 Flash exits preview
With improved performance and efficiency, Gemini 2.5 Flash now ranks just behind 2.5 Pro on the LM Arena leaderboard.

Gemini 2.5 Deep Think announced
This new mode specialises in mathematics, coding, and multimodal tasks. It will roll out to Ultra subscribers after further safety evaluations.

Imagen 4 and Veo 3 arrive
Imagen 4 delivers sharper image and text output inside the Gemini app. Veo 3 adds native audio generation to video creation, available with Ultra access in the US.

Gemini Live gets new features
Camera and screen sharing are coming to the iOS app. Calendar, Maps, Keep, and Tasks integration will follow soon.

Deep Research gets PDF and image uploads
Gemini now lets users upload files to aid Deep Research, combining personal inputs with public information for better answers.

Canvas turns your text into multiple formats
Users can now convert Gemini Canvas content into webpages, infographics, quizzes, or even audio overviews.

Gemini and Gemini Live land on desktop Chrome
Both Mac and Windows users will soon be able to ask questions about the page they’re viewing, with full navigation support expected later this year.

AI Mode comes to Google Search
Search now features a dedicated AI Mode for deeper insights. It integrates with Gmail for personalised results and can create graphs for sports and finance queries.

Agent Mode enables AI task automation
Using Project Mariner, Gemini's Agent Mode can book tickets, make reservations, or even buy products via Google Pay, pending user approval.

Gmail, Docs, and Vids get AI upgrades
Gmail introduces smart replies and inbox cleanup. Google Docs integrates Imagen 4 for visuals, while Vids transforms presentations into videos using AI avatars and scripts.

Project Starline becomes Google Beam
Renamed as Beam, Google’s 3D video communication platform gets an AI-first focus, with HP set to release the first supported hardware.

Jules debuts as an AI coding agent
A new tool for developers, Jules can understand code, fix bugs, write tests, and upgrade dependencies automatically. It’s available globally in public beta with no waitlist.