The interesting part of Gemini Omni is not that you can change the colour of a car or replace a background. We can already edit things. The shift is how you ask for the change. Instead of rebuilding the prompt or manually isolating every element, you can respond to the video the way you might respond to an editor: keep that performance, lose the car, make it dusk, leave the camera move alone.
That could make AI video feel much less like generating clips and much more like reviewing work. You make something, react to it, give direction and keep refining. For creative teams, the interface starts resembling the language already used in edit suites, art direction sessions and client reviews.
