Multimodal2026-05-31
Google AI Blog
9 Demos of Gemini Omni and Gemini 3.5 in Action
Google has released nine new demonstration videos showcasing the remarkable capabilities of its latest AI models, Gemini Omni and Gemini 3.5, which were officially announced at Google I/O 2026. The demos provide a hands-on look at how these models are pushing the boundaries of multimodal artificial intelligence.
The videos highlight several key advancements that set Gemini Omni and Gemini 3.5 apart from their predecessors. One of the most impressive demonstrations involves real-time video understanding. In the demo, the model watches a live video feed of a person assembling a piece of furniture and provides step-by-step verbal guidance, correcting mistakes and answering questions about the process as they occur.
Another demo focuses on complex reasoning across different data types. A user shows the model a hand-drawn sketch of a business process, uploads a related spreadsheet, and asks for a written analysis. Gemini Omni seamlessly integrates the visual information from the sketch with the numerical data from the spreadsheet to produce a coherent, insightful report.
The demos also showcase enhanced real-time interaction capabilities. Unlike previous models that required a pause between input and output, Gemini 3.5 demonstrates near-instantaneous conversational flow, complete with the ability to interrupt, ask clarifying questions, and adjust its tone based on user feedback. This makes interactions feel more natural and human-like.
Other demonstrations include advanced code generation from whiteboard diagrams, real-time language translation with contextual awareness, and the ability to analyze long-form video content, such as a full lecture, and generate a detailed summary with timestamps. These videos collectively paint a picture of an AI ecosystem that is becoming more integrated, intuitive, and capable of handling the messy, multimodal nature of real-world problems. Google has made the full playlist available on its official YouTube channel for developers and enthusiasts to explore.