Google I/O 2025: Veo 3, Imagen 4 and Gemini's New Paid Plans

Google held its annual I/O conference today and used the stage to roll out the biggest update to the Gemini app since its launch. The company introduced a new video model with native sound, a revamped image generator, free features for all Gemini Live users, and, for the first time, a two-tier paid subscription structure.

Veo 3, the first model to generate video with sound and dialogue

The centerpiece of the announcement is Veo 3, Google's new video generation model. According to the company, it's "the first in the world to have native support for sound effects, background noises and dialogue between characters." Until now, AI video generation meant producing moving images only; any audio was added afterward using separate tools. Veo 3 builds sound directly into the generation process: from a simple text prompt, the model creates both the scene and its accompanying soundscape, from the din of a city street to the rustle of leaves or a conversation between the characters on screen.

Availability, however, is limited for now: Veo 3 is live today in the Gemini app exclusively for Google AI Ultra subscribers in the United States, the priciest tier of Google's new lineup.

Alongside Veo 3 comes Imagen 4, the latest generation of Google's image model, which the company describes as faster and better at rendering text and typography within images — a historically weak spot for AI image generators. Unlike Veo 3, Imagen 4 is available today to all Gemini app users, with no plan restrictions.

Gemini Live, free for everyone with camera and screen sharing

Google also dropped the paywall on one of its most talked-about features: Gemini Live, the conversational mode that lets users share their phone's camera or screen in real time so the AI can see what they see and help solve a problem on the spot. As of today, this feature is free on Android and iOS for everyone, no subscription required.

Google says conversations in Gemini Live last, on average, five times longer than text-based ones, which the company attributes to the new ways it lets people ask for help — from troubleshooting a broken appliance to getting personalized shopping advice. In the coming weeks, Gemini Live will integrate more deeply with Google's ecosystem: it will be able to create Google Calendar events from a conversation about plans with friends, or check Google Maps if a user asks about a restaurant. The company also announced future connections with Tasks and Keep, manageable from the app's settings.

Deep Research, Canvas and Gemini's arrival in Chrome

Deep Research, the tool built for compiling research reports, now lets users add their own sources — such as PDFs or images — to cross-reference with public data. Google gives the example of a market analyst uploading internal sales figures in PDF form to compare against public trends, or an academic pulling in hard-to-find journal articles to enrich a literature review. The company says users will soon be able to research directly across Google Drive and Gmail as well.

Canvas, the creative space within the Gemini app, now benefits from the Gemini 2.5 models, enabling interactive infographics, quizzes and "Audio Overviews" (podcast-style summaries) in 45 languages. Google highlights 2.5 Pro's ability to translate complex ideas into working code with remarkable speed, which the company links to the "vibe coding" trend: describing an app in words and watching it take shape, lowering the barrier to building software.

Google also announced that Gemini is coming to Chrome. Starting tomorrow, it will begin rolling out on desktop to Google AI Pro and Google AI Ultra subscribers in the U.S. who use the browser in English on Windows and macOS. This first version lets users ask Gemini to clarify complex information on any webpage they're reading or summarize it. Google says that, down the road, the AI will be able to work across multiple tabs and browse websites on the user's behalf.

Interactive quizzes and support for students

Among the education-focused updates, Gemini now offers interactive quizzes: users can simply ask it to, for example, create a practice exam on thermodynamics to get a personalized learning experience, with instant feedback that flags topics needing more attention. Once finished, Gemini proactively offers a follow-up quiz focused on those weak spots. The feature is rolling out today to all Gemini users worldwide, on both desktop and mobile.

Additionally, college students in the United States, Brazil, Indonesia, Japan and the United Kingdom are eligible for a free full school year of the Google AI Pro plan, with more countries expected soon.

Two paid plans: Google AI Pro and Google AI Ultra

The most structural announcement of the day was the launch of two new subscriptions that replace and expand Google's previous offering. Google AI Pro costs $19.99 a month, replaces and broadens Gemini Advanced, and includes tools like Flow and NotebookLM, with special features and higher usage limits.

Google AI Ultra, meanwhile, is the top tier: access to Google's most powerful models with the highest usage limits and early access to the most notable experimental products before anyone else, including Veo 3 and the upcoming 2.5 Pro Deep Think mode. Ultra subscribers also get early access to Agent Mode, a new experimental capability coming soon to desktop: according to Google, users will simply need to state a goal, and Gemini will orchestrate the necessary steps on its own, combining live web browsing, in-depth research and integrations with Google apps to handle complex, multi-step tasks with minimal user oversight.

The Ultra plan costs $249.99 a month and, for now, is only available in the United States, though Google says it will expand to more countries. New users get a 50% discount for the first three months.

Google also confirmed that 2.5 Flash is now the default model in the Gemini app, combining, in the company's words, high quality with very fast response times.

What changes for current Gemini users

The announcements point to two clear moves. On one hand, Google is broadly giving away features that were previously paid or experimental, such as Gemini Live with camera and screen sharing or the interactive quizzes, instantly widening the pool of users who can try these capabilities. On the other, it's reserving the most computationally expensive features, like Veo 3 or the upcoming Agent Mode, for an Ultra plan that runs close to $250 a month — a figure that places Google at the high end of the AI subscription market and marks a striking gap versus its Pro plan, which is nearly twelve times cheaper.

Gemini's integration into Chrome and its growing ties to Calendar, Maps, Tasks and Keep also point to a goal the company has stated outright: turning Gemini into an assistant that doesn't just answer questions, but takes action within the app ecosystem hundreds of millions of people already use every day.