Summary
The AI Search channel breaks down Google’s 2026 I/O announcements, covering a dense set of new AI products and model updates in plain terms. The headline addition is Gemini Omni, Google’s new multimodal video model capable of accepting text, images, video, and audio inputs to generate or edit video — including background replacement, object removal, camera angle changes, and audio-synchronized effects. The presenter offers an early comparative take, finding Gemini Omni trailing Seed Dance 2.0 on anatomy realism and high-action sequences, while praising its educational explainer video output.
Also launched is Gemini 3.5 Flash, Google’s new near-frontier model targeting agentic workflows. Google claims it runs four times faster than competing frontier models on output tokens per second and supports multi-agent architectures where complex tasks are split across parallel AI workers — relevant for code generation, document organization, and multi-step tool use. Gemini Spark is introduced as a 24/7 personal agent built on Gemini 3.5 with deep Google Workspace integration (Gmail, Docs, Slides), running persistently in the cloud. Google says Spark is rolling out to trusted testers immediately, with a US Ultra subscriber beta planned for the following week.
Rounding out the announcements is a significant Google Search redesign that moves the product away from link retrieval toward a conversational AI interface. New background information agents monitor the web continuously for user-specified topics, and agentic booking capabilities let Search surface pricing, availability, and direct booking links without leaving the results page.
📺 Source: AI Search · Published May 21, 2026
🏷️ Format: News Analysis
