Google I/O 2026 Unveils Gemini Omni, Spark, and Antigravity 2.0
Google is positioning generative AI not as a chat feature but as an infrastructure layer embedded across search, development tools, and wearable hardware, with the new smart glasses marking a concrete consumer product for the Gemini ecosystem.
Reporting from 1 sources: ASCII.jp.
At its Google I/O 2026 developer conference on May 19, Google announced a suite of new AI features and products centered on evolving AI from a question-answering tool into an agent that performs tasks. The highlight is Gemini Omni, a new model described as the video version of the existing Nano Banana. Google DeepMind CTO Koray Kavukcuoglu said it can generate video from any combination of text, images, audio, and video input, and supports conversational video editing such as changing lighting or materials while maintaining character consistency. In the Gemini app, a personal AI agent called Gemini Spark was announced. It operates based on user instructions, integrates with apps the user chooses, and asks for confirmation before spending money or sending emails. Spark rolls out to trusted testers this week, with a beta for US Google AI Ultra subscribers next week. For developers, Google Antigravity 2.0 was announced as a standalone desktop app for running multiple AI agents in parallel. It includes dynamic sub-agents, scheduled tasks, and integration with Google AI Studio, Android, and Firebase. Google also announced new smart glasses compatible with Gemini, launching in fall 2026 in voice-help and information-display types, using the Android XR platform co-developed with Samsung and Qualcomm, with frames from Gentle Monster and Warby Parker.
Gemini Omni, the flagship model announced at Google I/O 2026, is described by Google DeepMind CTO Koray Kavukcuoglu as a model that can generate video from any input, combining text, images, audio, and video. It can edit existing videos through conversation alone, such as dimming lighting or changing a sculpture's material, while preserving character consistency and physical naturalness. The model will also be available in the AI creation tool Google Flow for creators.
The personal agent Gemini Spark operates based on user instructions and integrates with apps the user chooses. For sensitive operations like spending money or sending emails, it asks for confirmation before executing. Spark will roll out to trusted testers this week, with a beta version for US Google AI Ultra subscribers next week. MCP connections with Canva and others start today, and Spark will be able to use tools in the coming weeks. Planned features include requesting tasks via text or email, creating custom sub-agents, and local browser operations. The macOS version of the Gemini app will introduce Spark this summer, supporting tasks using local files.
Google Antigravity 2.0 is a standalone desktop app for running multiple AI agents in parallel. It includes dynamic sub-agents, scheduled tasks for background automation, and integration with Google AI Studio, Android, and Firebase. Related tools include Antigravity CLI for terminal use, Antigravity SDK for programmatic agent functionality, and integration with the Gemini Enterprise Agent Platform. Managed Agents were introduced to the Gemini API, allowing a single API call to launch an agent that reasons, uses tools, and executes code in an isolated Linux environment.
Google Search is also being agentified. For US Google AI Pro and Ultra subscribers first, users can create custom mini-app-like experiences within Search using Antigravity, such as a health tracker incorporating real-time reviews, maps, local data, and weather. The generative UI feature will be free for all users this summer.
On the hardware side, Google announced new smart glasses compatible with Gemini, launching in fall 2026. Two types are planned: a voice-help type that responds only with audio, and an information-display type that shows information on a screen. The glasses use the Android XR platform co-developed with Samsung and Qualcomm, with frames from Gentle Monster and Warby Parker. They support basic tasks like asking questions about objects, navigation, taking and editing photos and videos, and translation, as well as multi-step task assistance using Gemini Intelligence. The assistant is activated via voice command or by tapping the side of the frame. The glasses will be compatible with both iPhone and Android smartphones.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.
Sources
- ASCII.jp Googleのスマートグラス、秋発売へ