Google has concluded its keynote address at I/O 2025, showcasing a range of announcements primarily focused on artificial intelligence. Attendees witnessed a series of updates across the company’s image and video generation technologies, along with new features in services such as Search and Gmail.
Among the highlights were unexpected revelations, including a new AI filmmaking application and enhancements to Project Starline. For those who missed the live broadcast, a comprehensive recap is available below.
Google’s AI Mode for Search is now accessible to all users in the United States. The company has begun deploying AI Mode, a feature that allows users to conduct web searches via the Gemini AI chatbot. This summer, Google plans to test additional functionalities within AI Mode, such as deep search capabilities and a tool for generating charts related to finance and sports queries. Users can also anticipate a shopping feature in AI Mode launching in the coming months.
Project Starline has been rebranded as Google Beam, marking a significant development in its evolution from a 3D video chat booth. The new iteration will be introduced within an HP device equipped with a light field display and six cameras, enabling the creation of a three-dimensional image of the person on the other end of a video call. Companies like Deloitte, Duolingo, and Salesforce have already committed to integrating HP’s Google Beam devices into their workspaces.
The company also introduced Imagen 4, an advanced version of its AI text-to-image generator, which boasts improved capabilities for generating text and new options for exporting images in various formats, including square and landscape. Additionally, Veo 3, the latest iteration of Google’s AI video generator, will allow simultaneous generation of video and audio. Meanwhile, Veo 2 has been enhanced with features such as camera controls and object removal tools.
In addition to these updates, Google launched a new AI filmmaking app named Flow. This tool leverages Veo, Imagen, and Gemini to produce eight-second AI-generated video clips based on user input through text prompts or images. Flow also features scene-building tools to help combine clips into longer videos.
Gemini 2.5 Pro has introduced an enhanced reasoning mode termed Deep Think, which is designed for tackling complex queries in math and coding. This experimental mode is expected to be available initially to a select group of trusted testers. Alongside this, Google has made the Gemini 2.5 Flash model accessible to users through the Gemini app and is implementing improvements to the cost-effective model in Google AI Studio prior to a broader rollout.
As part of a collaboration with Xreal, Google presented a prototype of Project Aura—an innovative pair of smart glasses utilizing the Android XR platform for mixed-reality experiences. Although details are limited, it was noted that the glasses will feature Gemini integration, an expansive field of view, and built-in cameras and microphones.
Google’s experimental AI assistant, Project Astra, is evolving to be more proactive. The latest prototype allows it to utilize the phone’s camera to recognize objects and complete tasks independently, such as identifying errors in homework without explicit instructions from the user.
Gemini is also set to be integrated into Chrome. Starting May 21st, subscribers to Google AI Pro and Ultra will have the option to utilize the Gemini button in Chrome for clarification or summarization of information across different webpages, with plans to enable multi-tab functionality later this year.
In a momentous move, Google announced the introduction of a new “AI Ultra” subscription, priced at $250 per month. This plan will provide customers with access to the company’s most advanced AI models and increased usage limits for applications such as Gemini, NotebookLM, and Flow, in addition to offering early access to Gemini in Chrome and a new capacity to complete up to ten tasks simultaneously via Project Mariner.
In tandem with Project Astra, Google revealed a feature called Search Live. This innovative tool allows users to engage in a back-and-forth conversation with the Search assistant while showcasing their camera’s view. Following the recent rollout of Gemini Live’s screen-sharing functionality to Android users, iOS users will also now be able to access it for free.
Google unveiled a new AI-driven application called Stitch, designed to automate the generation of user interfaces based on selected themes and descriptive input. Users can enhance Stitch’s output using wireframes, initial sketches, or screenshots of existing UI designs. Currently, this experiment is available through Google Labs.
Additionally, Google Meet is launching a feature that translates spoken language in real-time during conversations. Initially supporting only English and Spanish, this feature is available in beta for Google AI Pro and Ultra subscribers.
Upgrades are also coming to Gmail, with the smart reply function now incorporating data from users’ inboxes and Google Drive to generate responses that better reflect individual communication styles. This upgraded feature will also consider the tone of the recipient, enabling more contextually appropriate responses, particularly in formal settings.
As part of its commitment to enhancing AI in shopping experiences, Google is testing a feature that enables users to upload full-length images of themselves to virtually try on clothing items, supported by an AI model designed to understand body types and clothing nuances. Additionally, the anticipated shopping capabilities in AI Mode will also include the introduction of an “agentic checkout” feature, allowing the AI to make purchases on users’ behalf.
Finally, Google Chrome is set to improve user security by introducing a feature that alerts users to compromised passwords. When such a situation is detected, Chrome will be able to generate a strong replacement password and automatically update it on supported websites, all while ensuring user consent prior to making any changes.
