Google I/O 2025: Key Announcements & Highlights

Google I/O 2025: Key Announcements

The annual Google I/O developer conference is currently underway, taking place on Tuesday and Wednesday at the Shoreline Amphitheatre in Mountain View. Our team is reporting live from the event with the most recent news.

I/O serves as a platform for Google to unveil product innovations spanning its entire range of offerings. Expect significant updates concerning Android, Chrome, Google Search, YouTube, and notably, Google’s Gemini AI chatbot.

Android Updates Highlighted at The Android Show

Prior to the main I/O event, Google held a dedicated presentation, The Android Show, focusing specifically on Android advancements. Several new features were revealed.

These include enhanced methods for locating misplaced Android phones and other tagged possessions. Further device-level functionalities were also introduced for the Advanced Protection program.

Security enhancements designed to safeguard users from fraudulent activities and device theft were detailed. A fresh design aesthetic, known as Material 3 Expressive, was also presented.

Comprehensive Coverage of Google I/O 2025

Below is a summary of all announcements made during Google I/O 2025.

• New features for locating lost devices were demonstrated.
• The Advanced Protection program received expanded device-level capabilities.
• Enhanced security tools were unveiled to combat scams and theft.
• Material 3 Expressive, a new design language, was introduced.

Gemini Ultra

According to Google, Gemini Ultra represents the most comprehensive tier of access to their suite of AI-driven applications and services. Currently available solely within the United States, it is offered at a monthly subscription cost of $249.99.

This premium plan incorporates Google’s Veo 3 video generation tool, alongside the newly introduced Flow video editing application.

Furthermore, subscribers gain access to Gemini 2.5 Pro’s Deep Think mode, a potent AI feature that is scheduled for future release.

Enhanced Capabilities & Included Services

Gemini Ultra provides elevated usage limits within both NotebookLM and Whisk, Google’s platform for image manipulation.

Subscribers also benefit from integrated access to the Gemini chatbot directly within the Chrome browser.

Access is granted to several “agentic” functionalities, leveraging the capabilities of Google’s Project Mariner technology.

Additional Benefits

• A subscription to YouTube Premium is included.
• Subscribers receive a substantial 30TB of combined storage space distributed across Google Drive, Google Photos, and Gmail.

The plan is designed to offer a complete and powerful AI experience for users requiring advanced features and substantial resource allocation.

Deep Think in Gemini 2.5 Pro

Deep Think represents an advanced reasoning capability integrated into Google’s leading Gemini 2.5 Pro model.

This feature enables the model to evaluate a range of potential responses to prompts, ultimately leading to improved performance on specific evaluation metrics.

How Deep Think Functions

While Google has not disclosed the precise mechanisms behind Deep Think, its operation may share similarities with OpenAI’s o1-pro and the forthcoming o3-pro models.

These models likely employ a system designed to identify and combine the most effective solutions to complex queries.

Availability and Safety

Currently, Deep Think is being offered to a select group of “trusted testers” through the Gemini API.

Google is prioritizing thorough safety assessments before making Deep Think broadly accessible to all users.

Additional time is being allocated to ensure responsible deployment of this enhanced reasoning mode.

Key Benefits

• Improved reasoning capabilities.
• Enhanced performance on benchmarks.
• Multiple answer consideration.

Gemini 2.5 Pro with Deep Think aims to provide more accurate and insightful responses.

Veo 3: Google’s Advanced Video Generation AI

Google has announced that its Veo 3 model is now capable of producing accompanying audio elements, including sound effects, ambient noises, and spoken dialogue, alongside the videos it generates.

According to Google, Veo 3 represents a significant advancement over its previous iteration, Veo 2, delivering enhanced video quality.

Availability and Access

Access to Veo 3 is being rolled out starting Tuesday within Google’s Gemini chatbot application.

The feature is exclusively available to subscribers of Google’s AI Ultra plan, which is priced at $249.99 per month.

Users can initiate video creation by providing either text-based prompts or uploading an image as a starting point.

Key Capabilities

• Audio Generation: Veo 3 can synthesize relevant sound effects and background audio.
• Dialogue Creation: The model is able to generate spoken lines to complement the visual content.
• Improved Quality: A noticeable upgrade in video fidelity compared to Veo 2.
• Prompting Options: Videos can be created from text descriptions or image inputs.

This new functionality expands the creative possibilities offered by AI-powered video generation, allowing for more immersive and complete content creation.

Imagen 4 AI Image Generator

Google has announced that Imagen 4 is a significant advancement in speed compared to its predecessor, Imagen 3. Further optimizations are planned, with a future version projected to be up to ten times faster than Imagen 3.

The capabilities of Imagen 4 extend to the rendering of intricate details. These include textures like fabrics, the realism of water droplets, and the complexity of animal fur, as stated by Google.

Versatility in Style and Resolution

Imagen 4 demonstrates adaptability, successfully producing images in both photorealistic and abstract artistic styles. It supports a diverse range of aspect ratios and can generate images with resolutions up to 2K.

Both Veo 3 and Imagen 4 are slated to be integral components of Flow, Google’s AI-driven video platform designed for professional filmmaking applications.

The integration of these AI models into Flow will empower filmmakers with advanced tools for video creation and editing.

Google’s commitment to enhancing AI image generation is evident in the rapid development and release of these new models.

Gemini Application Updates and Growth

Google has recently reported that the Gemini applications have surpassed 400 million monthly active users globally.

This signifies substantial adoption of the AI-powered tools since their launch.

Enhanced Functionality with Gemini Live

This week marks the widespread release of camera and screen-sharing features within Gemini Live for both iOS and Android platforms.

Leveraging the technology behind Project Astra, these capabilities enable users to engage in almost instantaneous verbal interactions with Gemini.

Simultaneously, the AI model receives a live video feed from the user’s smartphone camera or screen.

Expanding Integration with Google Services

In the near future, Gemini Live will experience a more seamless integration with other prominent Google applications.

Users will soon be able to utilize Gemini to obtain directions directly from Google Maps.

Furthermore, the AI will facilitate event creation within Google Calendar and the generation of to-do lists using Google Tasks.

Deep Research Capabilities Improved

Deep Research, Gemini’s AI agent designed for in-depth research report creation, is also receiving updates.

A key enhancement allows users to upload and analyze their own confidential PDF documents and image files.

This feature expands the scope of research that can be conducted using the AI agent.

Stitch: AI-Powered UI Design

Stitch represents an artificial intelligence solution designed to assist in the creation of front-end designs for both web and mobile applications.

The tool functions by generating the required user interface (UI) components and associated code.

Designs can be initiated within Stitch through simple text prompts or even by uploading an image, resulting in the production of HTML and CSS markup.

Functionality and Scope

While offering a substantial degree of customization, Stitch’s capabilities are somewhat narrower in scope when contrasted with certain other AI-driven coding platforms.

It excels in UI generation, but may not encompass the full breadth of features found in more comprehensive tools.

Jules: AI Assistance for Developers

In addition to Stitch, Google has broadened access to Jules, an AI agent specifically engineered to aid developers in debugging code.

Jules is capable of assisting with the comprehension of intricate codebases.

Furthermore, it can facilitate the creation of pull requests on platforms like GitHub and manage specific tasks within a development backlog.

The tool is intended to streamline programming workflows and enhance developer productivity.

Project Mariner

Project Mariner represents Google’s innovative foray into the realm of artificial intelligence agents capable of web browsing and utilization.

Google has announced substantial enhancements to the operational mechanisms of Project Mariner, enabling the agent to concurrently manage approximately twelve different tasks.

A wider release of this updated functionality is currently underway, making it accessible to a growing user base.

Key Capabilities

The functionality of Project Mariner allows users to accomplish tasks such as securing tickets for a baseball game or completing online grocery shopping.

Notably, these actions are performed entirely through interaction with Google’s AI agent, eliminating the need for direct navigation to external websites.

Users can simply engage in conversational dialogue with the AI, and it will autonomously visit websites and execute the requested actions on their behalf.

How it Works

• Project Mariner functions as an AI agent.
• It autonomously browses the web.
• It interacts with websites to complete tasks.
• Users interact via a conversational interface.

This streamlined process offers a convenient and efficient alternative to traditional online task completion methods.

The agent’s ability to handle multiple tasks simultaneously further enhances its utility and user experience.

Project Astra

Project Astra represents Google's innovative, low-latency, and multimodal AI experience.

This technology is poised to enhance a variety of applications, including Google Search, the Gemini AI app, and offerings from external developers.

Origins and Development

Initially developed within Google DeepMind, Project Astra was conceived as a demonstration of advanced, near real-time multimodal AI functionality.

Currently, Google is collaborating with partners such as Samsung and Warby Parker to develop dedicated Project Astra glasses.

However, a definitive launch date for these glasses has not yet been established.

The project highlights Google’s commitment to pushing the boundaries of AI and its practical applications in everyday life.

A New Search Experience: AI Mode

This week marks the beginning of a rollout to users within the United States for AI Mode, an innovative and experimental feature from Google Search.

This new functionality empowers individuals to pose intricate, multi-faceted questions through an artificial intelligence-driven interface.

Capabilities and Features

AI Mode is designed to handle sophisticated data within the realms of both sports and finance-related inquiries.

Furthermore, it introduces an interactive "try it on" capability specifically for clothing and apparel searches.

Introducing Search Live

Expanding the scope of AI-powered search, Search Live is scheduled for release later this summer.

This feature will enable users to formulate questions based on the immediate visual input captured by their smartphone’s camera.

Personalized Context in Gmail

Google is integrating personalized context into its applications, beginning with Gmail.

Gmail is now the first application to benefit from this enhanced contextual awareness, improving the user experience.

Beam 3D Teleconferencing

Beam, formerly known as Starline, employs a sophisticated integration of software and specialized hardware. This includes a six-camera system and a unique light field display, designed to facilitate conversations that feel as natural as being physically present in the same room.

An artificial intelligence model processes video feeds captured by the array of cameras. These cameras are strategically positioned at various angles, focusing on the user, and the AI transforms this data into a detailed 3D representation.

Key Features and Capabilities

Google’s Beam system is characterized by its exceptional precision, offering what is described as “near-perfect” millimeter-level head tracking. It also delivers video streaming at a smooth 60 frames per second.

When integrated with Google Meet, Beam unlocks an AI-driven, real-time speech translation capability. This feature is notable for its ability to maintain the nuances of the original speaker’s voice, including tone and emotional expression.

Integration with Google Meet

Further enhancing the communication experience, Google has announced the addition of real-time speech translation directly within Google Meet.

This new functionality expands the accessibility and global reach of meetings conducted through the platform.

Recent Advancements in Artificial Intelligence

Google is integrating Gemini into the Chrome browser, introducing a novel AI-powered browsing assistant.

This new tool is designed to facilitate rapid comprehension of webpage content and streamline task completion for users.

Gemma 3n, a model optimized for performance across various devices, has been unveiled.

It is engineered to operate efficiently on smartphones, laptops, and tablets, and is currently available in preview.

Gemma 3n demonstrates versatility by processing multiple data types, including audio, text, images, and video, as stated by Google.

AI Enhancements to Google Workspace

A significant number of AI-driven features are being introduced to the Google Workspace suite, impacting applications like Gmail, Google Docs, and Google Vids.

Gmail will benefit from personalized smart reply suggestions and a new functionality for inbox management.

Google Vids is receiving expanded capabilities for content creation and editing.

Further AI Innovations from Google

Video Overviews are being integrated into NotebookLM, enhancing its functionality.

The SynthID Detector, a verification portal utilizing Google’s SynthID watermarking technology, has been launched.

This tool assists in the identification of content generated by artificial intelligence.

Lyria RealTime, the AI model underpinning Google’s experimental music production application, is now accessible through an API.

This allows developers to integrate its capabilities into their own projects.

Wear OS 6

A more consistent visual experience is arriving with Wear OS 6, featuring a standardized font across all tiles. This update aims to provide a more polished and unified appearance for applications.

Specifically for Pixel Watches, the new version introduces dynamic theming. This feature allows app colors to intelligently adapt and harmonize with the currently selected watch face.

Enhanced Customization for Developers

The foundation of Wear OS 6 is a new design reference platform. This platform empowers developers to create more extensive customization options within their applications.

A key goal is to facilitate smoother and more fluid transitions between different app elements and the overall watch interface.

Resources for Implementation

To support developers in adopting these changes, Google is providing comprehensive design guidelines.

These resources are accompanied by readily available Figma design files, streamlining the integration process.

Google Play Store Updates for Developers

Google is enhancing the Play Store platform for Android developers with a suite of new functionalities. These improvements focus on streamlining subscription management, enabling focused content discovery, and simplifying the add-on sales process.

New topic pages are being introduced, allowing users to explore content centered around specific interests. Initially available in the U.S., these pages will link users to relevant applications associated with various movies and television programs.

Enhanced Developer Tools

Developers are now provided with dedicated spaces for testing and managing app releases. This includes tools for monitoring and optimizing the rollout of their applications.

A crucial feature allows developers to pause live app releases immediately should any critical issues arise during deployment.

Streamlined Subscription Options

The subscription management system is receiving significant updates, notably with the introduction of multi-product checkout.

Soon, developers will have the capability to offer supplementary subscription add-ons in conjunction with primary subscriptions, all processed through a single transaction.

Furthermore, a new checkout experience is being implemented to facilitate a more seamless process for selling application add-ons.

Audio samples are also being introduced, providing users with a preview of in-app audio content before downloading.

Android Studio

New AI features are being incorporated into Android Studio, notably “Journeys.” This represents an “agentic AI” functionality, launched alongside the Gemini 2.5 Pro model.

Furthermore, an “Agent Mode” is slated for implementation, designed to manage more complex development workflows.

Enhanced Capabilities

Android Studio is set to benefit from expanded AI-driven features, including a refined “crash insights” tool within the App Quality Insights panel.

Leveraging the power of Gemini, this enhancement will scrutinize an application’s code base to pinpoint likely origins of crashes and propose resolutions.

App Quality Improvements

The improved crash insights feature aims to provide developers with a more efficient method for debugging and stabilizing their applications.

By automatically analyzing source code, the tool reduces the time required to identify and address critical issues.

Gemini Integration

The integration of Gemini 2.5 Pro is central to these advancements, providing the necessary processing power for sophisticated code analysis.

This allows Android Studio to offer more intelligent and proactive assistance to developers.