Google I/O 2025: AI, XR, Android and other announcements

Sundar Pichai is wrapping up I/O by revealing that they are building a global network of satellites to detect fires.

Gentle Monster and Warby Parker will be the first eyewear partners to build Android XR glasses.

Google describes Android XR as the “first Android platform built for the Gemini era.” Developed in collaboration with Samsung and Qualcomm, Android XR powers Samsung’s upcoming Project Moohan headset—which bears a striking resemblance to Apple’s Vision Pro.

Google has introduced two new AI subscription plans: Google AI Pro and the all-new Google AI Ultra. The Pro tier is priced at $19.99 per month, while the Ultra plan comes in at a steep $249.99 per month—and yes, that’s not a typo.

Google is launching Flow, a new creative tool for artists that combines the capabilities of its Veo video generator, Imagen image model, and Gemini AI chatbot. Flow allows users to prompt with precise camera controls and seamlessly extend scenes. Beyond video trimming, it also offers the ability to lengthen existing clips.

Google highlights SynthID, its technology that has embedded watermarks in over 10 billion images. The company states that SynthID is becoming easier to implement across more media types and simpler to detect.

Google has also announced Veo 3, its enhanced video generator, boasting improved visual quality and physics understanding. A significant advancement is its "native audio generation," enabling the model to generate spoken dialogue for characters and incorporate background sounds and music directly into the videos.

Google has announced a new Imagen 4 image generator, available in Gemini. You can give text prompts to create images—similar to ChatGPT. It remains to be seen if Imagen 4 in Gemini will kick-off another Ghibli trend.

Gemini is integrating into Chrome as an AI assistant that understands the context of the webpage you are viewing. This feature will begin rolling out this week to Gemini subscribers in the US.

Gemini Live is getting support for camera and screen sharing starting today for Android and iOS apps.

Gemini is going to be a "personal, proactive, and powerful" assistant. You can soon add even more personal context from across Google services to Gemini, including hand-written notes from a lecture. Deep Research in Gemini will let you upload your own files to help guide it.

Google Search's new AI Mode is also gaining Deep Search as well as agentic capabilities from Project Mariner. This seems like the future of Google Search. AI Mode will also offer personalized shopping capabilities, including virtual try-ons for clothes and an AI assistant that can complete purchases on your behalf.

Google Search is getting more intelligent, agentic, and personalized, as per Sundar Pinchai. Google says it has completely reimagined Search with AI Mode, allowing users to ask much longer and more complex questions, a trend already evident with 2-3x longer queries. This new experience is available starting today as a separate tab within Search for all users in the US. AI Mode will have all of Google's best AI features and capabiltiies.

Google just showed a "concept video" of Project Astra wherein Gemini Live helped a user repair a bike. The AI chatbot starts by finding the instruction manual in PDF form, then pulls up a relevant YouTube tutorial. Next, it searches through the user’s emails for past correspondence with the bike shop. Finally, it places a call to the shop to order replacement parts while referencing the manual—all within a single, continuous workflow.

Gemini is going to be a "world model" that interacts with the physical environment like the human brain, says Demis Hasabis, Google DeepMind's CEO.

Google has announced Gemini 2.5 Flash, a faster and more affordable AI model. It surpasses previous models in almost every aspect and ranks second on LLMArena, just behind 2.5 Pro, the flagship variant. Gemini 2.5 Flash will be available in early June. Both 2.5 Pro and Flash also feature native audio output with two voice options, enabling more expressive conversations in multiple languages.

Gmail is getting personalized smart replies that can tap into your emails, notes, and Drive to craft responses based on your past conversations and personal context. "It can make you a better friend," says Sundar Pichai, noting that the feature helps users send thoughtful replies to friends using their own data and history.

As part of Project Mariner, AI agents can interact and operate browsers as well as other software. Mariner can now manage up to 10 simultaneous tasks, and you can now show it a task once and it will learn it for future cases. Agent mode coming to Chrome, search, and the Gemini app.

Real-time speech translation is coming to Google Meet starting today. English to spanish translation is available now and more languages will be added in the coming weeks.

Google Beam announced. It is Google's 3D video conferencing system wherein six cameras capture you from different angles. Built-in AI renders you in a 3D-like display at 60fps real-time rendering. HP will release the first Google Beam devices later this year.