Google I/O 2025: Gemini 2.5, Project Astra and 10 Groundbreaking AI Updates

In one of its most ambitious showcases to date, Google pulled back the curtain on a sweeping array of AI innovations at its annual Google I/O 2025 conference. The headline acts? A turbocharged Gemini 2.5 AI model and the debut of Project Astra—a bold new initiative to help machines understand the physical world around them.

Sundar Pichai and top executives outlined how AI will power the next generation of Google's products and services. With an emphasis on practical functionality, inclusivity, and cultural awareness, these developments aim to make AI more intuitive and useful than ever. Here's a breakdown of the most significant reveals:

1. Google Beam: The Future of 3D Video Calling

Formerly known as Project Starline, Google's 3D communication system has rebranded as Google Beam. It uses six cameras to capture multiple angles of a user's video stream and merges them into a 3D light field display with AI.

With head-tracking sensors and 60fps precision, Beam is being rolled out in partnership with HP and will debut with select customers later this year. Additional hardware will appear at InfoComm 2025 in June.

2. Gemini 2.5: Smarter, Faster, Sharper

The new Gemini 2.5 series gets a major intelligence upgrade. A feature called Deep Think mode—currently in testing—enhances reasoning in the 2.5 Pro model.

Native Audio Output will allow Gemini to produce more expressive, human-like speech, accessible via the Live API. Meanwhile, the 2.5 Flash model is being optimised for improved performance and affordability. Developers will also benefit from tools like thought summaries and a thinking budget.

3. AI Mode in Search: Welcome to Smart Browsing

A standout announcement was AI Mode in Search, powered by a tailored Gemini 2.5 model. New additions include Deep Search, a Live Search feature that uses the device camera, and agentic functions for booking appointments or buying tickets.

Shopping is also getting a revamp. Users can visually search for products, try on clothes virtually, and let AI agents track prices or even make automatic purchases.

4. AI Overviews: Going Global

Google's AI-generated Overviews for search are now expanding to more than 200 countries and over 40 languages, including Arabic, Chinese, Malay, and Urdu. This builds on existing support for languages like English, Hindi, Japanese, Portuguese and Spanish.

5. Gemini-Powered Android XR: Smart Glasses Get Smarter

Gemini will now power Google's Android XR platform, which will run on Samsung's Project Moohan smart glasses. These glasses include a camera, microphone, speakers and a display—enabling hands-free interactions with Gemini.

Users will be able to take photos, control smart devices, and interact with their phones simply by speaking.

6. Imagen 4 and Veo 3: AI for Creators

Google introduced Imagen 4 (image generation) and Veo 3 (video creation), both boasting improved rendering, text handling and quality.

A new app called Flow combines these models with Gemini to generate eight-second clips that can be pieced together into longer scenes using text or images as prompts.

7. Gemini in Google Chrome: Smarter Surfing

Gemini is being integrated into Chrome for paid users. A new button will let users summarise pages or ask questions about content. It can also auto-navigate websites and work across multiple tabs.

8. Stitch Tool: Code-Free Interface Design

The new AI-powered Stitch tool generates app interfaces from plain text descriptions or rough visuals like wireframes and sketches. Currently in experimental release via Google Labs, it aims to simplify UI creation.

9. Real-Time Speech Translation in Google Meet

Google Meet will now feature real-time speech translation, starting with English and Spanish. This feature enables smoother conversations across language barriers and is currently in beta for paid subscribers.

10. New AI Subscriptions: Google AI and AI Ultra

Google unveiled two new subscription tiers. The standard Google AI plan (£14.92 or $19.99/month) replaces the Google One AI Premium plan. For £186.58 or $24.99/month, the AI Ultra plan offers early access to features, higher usage limits, and 30TB of cloud storage.

Across the board, Google's new AI features focus on elevating human interaction, improving accessibility, and aligning with cultural context. Rather than forcing change, these tools aim to blend seamlessly into everyday life—bringing AI closer to the world it's meant to serve.