Google unveils Gemini Omni

Google used its I/O developer conference on May 19 to introduce Gemini Omni, a new model family for generating and editing short videos from combinations of text, images, audio and video. Google said the first release, Gemini Omni Flash, is available starting now in the Gemini app, Google Flow and YouTube Shorts. The company described Omni as a system that can “create anything from any input,” though the launch is starting with video output rather than a full multimodal publishing stack. TechCrunch reported that the product lets users edit clips through conversation and supports style and audio references. ### What exactly did Google launch at I/O? Google’s official I/O posts said Gemini Omni is a new model family built to take mixed inputs — including text, images, audio and video — and turn them into generated or edited video. The first model, Gemini Omni Flash, is the initial public step in that family and is aimed at faster creation and editing tasks. Google said broader output modes, including image and audio, are planned later. (blog.google) Sundar Pichai said in Google’s keynote post that Omni Flash would be available “starting today” across the Gemini app, Flow and YouTube Shorts, with developer and enterprise API access to follow in the coming weeks. That ties the launch not only to Google’s consumer apps but also to creator and developer products. ### How does Omni Flash work in practice? (blog.google) TechCrunch reported on May 19 that Gemini Omni can generate and edit videos through natural-language conversation, rather than requiring users to rebuild a clip from scratch after each change. Google’s product post said users can refine scenes, make changes with follow-up prompts and combine multiple kinds of source material in one workflow. (blog.google) CNBC reported that Google presented Gemini Omni as part of a wider push to expand Gemini from a chatbot and model family into a broader AI platform. Google’s Gemini app update post separately said Omni can transform text, images and video prompts into cinematic video outputs, placing the model inside the company’s main assistant product rather than as a standalone video tool. (techcrunch.com) ### What are the limits of the first release? TechCrunch reported that Omni Flash is focused on quick generation and editing of clips up to about 10 seconds long. The same report said the model supports style transfer and audio-reference inputs, giving users a way to steer the look or sound of short outputs without frame-by-frame editing. (cnbc.com) Google’s own announcement was more cautious on specific limits, emphasizing rollout locations and the broader “Omni family” roadmap rather than detailed public specs. That suggests some of the early operating details — including clip length and feature boundaries — are being communicated first through launch demos and partner briefings rather than a full technical specification sheet. That is an inference based on the difference between Google’s posts and third-party launch coverage. (techcrunch.com) ### Where does this fit in Google’s broader AI push? Google announced Gemini Omni alongside Gemini 3.5 Flash and Gemini Spark at I/O 2026, according to Google’s keynote materials and CNBC’s event coverage. Google said Gemini 3.5 Flash is its next-generation fast model for agentic tasks, while Spark is a personal AI agent inside the Gemini app. Omni extends that push into media generation and editing. (blog.google) Google I/O 2026 is running May 19-20 at Shoreline Amphitheatre in Mountain View, California, and online, according to Google’s event page. Google said the next step for Omni Flash is a wider rollout to developers and enterprise customers through APIs in the coming weeks. (blog.google 1) (blog.google 2)