Google shows Gemini Omni and 3.5 in nine demos
Google presents nine demos of Gemini Omni and Gemini 3.5 Flash: the first creates and edits videos through conversation, while the second handles complex tasks and builds interfaces and agents. Several features are already coming to Gemini, Search and Google's tools, although some require specific subscriptions or will launch later.
Google has shown nine demos of Gemini Omni and Gemini 3.5 Flash, two models introduced during Google I/O 2026 with very different uses in mind: creating and editing videos through text instructions, and delegating complex tasks to AI agents.
Gemini Omni turns an idea into video
Gemini Omni can combine images, audio, video and text to generate videos based on Gemini's knowledge of the real world. You can then modify them through a conversation, without starting from scratch each time.
Editing works through successive instructions. You can ask it to change an object, the setting, the lighting or the style while keeping the characters and scene consistent.
In one demo, a sculpture transforms into a figure made of bubbles. In another, the model changes a room to create a floating glass sphere with a recursive scene inside, like a set of rooms repeating indefinitely.
You can also alter the action in an existing video: add characters, introduce objects or turn an ordinary moment into a scene you could never have filmed. The idea is for the original video to be only the starting point.
Gemini 3.5 Flash works step by step
Gemini 3.5 Flash is designed for tasks in which AI must plan, make decisions and complete several actions in sequence. Google presents it as a fast model with programming capabilities and support for extended tasks, known as long-horizon tasks.
In a demo with Antigravity, Google's tool for coordinating agents, the model organizes unstructured files by renaming and classifying them according to criteria that can change during the process. With several subagents working under supervision, it can also handle broader workflows and programming tasks.
Another example shows it generating different proposals for a payment interface in about 60 seconds inside AI Studio. It does not simply respond with text: it creates web experiences, graphics and different ways to present information.
Search starts creating custom tools
Google is also bringing Gemini 3.5 Flash to Search. Search can generate visual interfaces, simulations and small programs on the fly, tailored to the user's specific question.
If you ask about a visual concept, it can build an interactive explanation. For needs that last weeks, such as organizing a wedding or starting an exercise routine, it can create dashboards, trackers or mini-apps that you can use again later.
These generative interface features will reach all Search users at no cost this summer. Custom experiences created with Antigravity will begin rolling out in the coming months to Google AI Pro and Ultra subscribers in the United States.
Google is also preparing what it calls information agents. They will work in the background to monitor specific topics, such as new products from an athlete, and send updates with links so you can check the details and take action.
An agent connected to your apps
The other offering is Gemini Spark, a personal agent based on Gemini 3.5 and Antigravity. It can help you manage digital tasks under your direction and connect to Gmail, Docs, Slides and other Workspace tools.
In the demo, Spark prepares a list of nut-free appetizers and then adds them to Instacart. The example shows the main shift: AI does not just recommend something. It can also complete part of the process in other apps.
Gemini Spark is already available to Google AI Ultra subscribers in the United States. Information agents will first arrive on Pro and Ultra plans this summer.
What you can use now
- Gemini Omni Flash is being added to Gemini and Google Flow for Google AI Plus, Pro and Ultra subscribers worldwide.
- It is also coming to YouTube Shorts and the YouTube Create app at no cost.
- Google expects to offer it through an API to developers and businesses in the coming weeks.
- Gemini 3.5 Flash is available in Google Antigravity, the Gemini API through AI Studio, Android Studio, Gemini Enterprise Agent Platform and Gemini Enterprise.
- The model is rolling out globally in the Gemini app and is already available to everyone in Search's AI Mode.
For you, the most important change is not just that the models generate better answers. It is that Google is placing them inside tools that can edit content, create interfaces, monitor information and take action. The question to watch now is how much control and supervision these agents will need when they move from demos to real, ongoing tasks.