AI video is becoming less like a one-time request and more like a conversation you can steer.
Google's Gemini Omni 1.2 Flash can generate video clips up to 15 seconds long. It can also revise them through follow-up instructions. Google says it follows visual references more consistently, too.
That matters because a generated clip can be a starting point, not a final answer. A creator can describe what should change and keep refining the result.
Think of giving notes to a director: see a revision, then give another note. The model can respond across multiple turns, instead of treating each prompt as a separate request.
The model is generally available through Google's developer API and is rolling out in Google Vids and Gemini for business. Google says video generation costs 30% less than its previous offering. Those routes bring the capability to developers and people making videos for work.
The bigger shift is treating generated video as a draft you can shape, not just an output to accept or reject. How well can a conversation guide a clip all the way to a finished video?
