Gemini Omni brings multimodal video generation to the forefront
Google CEO Sundar Pichai unveiled Gemini Omni at the I/O 2026 keynote on May 19. The new model accepts text, images, and video as inputs and can generate high-quality video grounded in real-world knowledge. This marks a departure from previous models that required separate tools for different media types.
Gemini Omni builds on the capabilities of Gemini 3.5, which was also announced at the conference. The model can understand a video clip, read accompanying text, and generate a new video that combines both sources of information. Google said Omni is designed for creators, educators, and businesses that need to produce video content quickly.
The model is available through Google AI Studio and will be integrated into YouTube and Google Workspace later this year. Pricing has not been announced, but Google said it will be competitive with other video generation services.
Gemini Spark runs 24/7 as a personal agent in the cloud
Gemini Spark is Google's first always-on personal AI agent. It lives in the cloud and runs continuously, performing tasks without waiting for user commands. Spark can monitor email, track packages, manage calendars, and alert users to changes in real time.
Spark is designed to handle background tasks that users normally check manually. For example, it can rebook a flight if a delay is detected, order groceries when supplies run low, or flag important emails while the user sleeps. Google said Spark operates across all Google services and third-party apps that integrate with the platform.
Privacy and security were a focus of the announcement. Spark runs in a secure cloud environment, and users control what data it can access. All agent actions are logged and can be reviewed. Google emphasized that Spark does not share data across user accounts.
Google shifts strategy toward agentic AI
The I/O 2026 announcements reflect Google's broader shift toward agentic AI. Instead of building chatbots that respond to queries, Google is building AI systems that act independently. The company said this approach will transform how users interact with technology.
Pichai called 2026 the year of the AI agent. He demonstrated Spark handling a complex travel booking that required coordinating flights, hotels, and restaurant reservations across multiple services. The agent completed the task in under a minute without human input at each step.
Google also announced updates to Gemini for Science, a program offering AI tools for researchers. The company is expanding AI Studio with more agent-building capabilities, allowing developers to create custom agents for specific tasks. Analysts say Google's agent strategy could reshape the competitive landscape against Microsoft's Copilot and OpenAI's ChatGPT.