Google launches Gemini Omni, a video model built around conversational editing

Google has launched Gemini Omni, a new family of video models. The first version, Gemini Omni Flash, takes images, audio, video, and text as inputs and produces video as output. The pitch is not the existence of a video model, which is by now a crowded category, but the way you edit one. Instead of cutting and masking through a timeline, you tell the model what to change in plain language, and successive instructions build on the previous result while keeping the scene consistent. Google also says Omni has a sharper grasp of physics, with better intuitive handling of gravity, kinetic energy, and fluid dynamics than past Gemini video work.

The release strategy is broader than usual. Omni Flash is rolling out to Google AI Plus, Pro, and Ultra subscribers through the Gemini app and Google Flow, and Google is putting it free on YouTube Shorts and the YouTube Create app, which means a very large audience will see it without ever opening a Gemini surface. Developer and enterprise APIs come in the following weeks. Every video gets a SynthID watermark, the same provenance scheme Google has been pushing across image and audio generation. Users can also produce videos starring digital avatars of themselves using their own voice.

The interesting test for Omni is not its launch demos but how well the conversational edit loop holds up after a dozen rounds, when scene drift usually starts to creep in.

Why it matters

If you ship video, the Omni Flash rollout puts a serious editor in front of consumers who do not own one today, and the YouTube Shorts and Create distribution means your viewers will start expecting iterative, prompt-driven changes before your tools catch up.