Google's new video model makes longer clips and lets you edit them in rounds

What Google released

Google has released a new generation of video generation models, as 36Kr reported. The model named in the coverage is Gemini Omni 1.2 Flash, an AI system that turns text instructions into video.

Its headline change is that a single clip can now run longer than before. The report as we have it gives no exact figure beyond that direction.

Editing without starting over

The second addition is multi-turn editing. That means you can refine a clip across several instructions, much like chatting with an assistant, instead of regenerating the whole video each time you want a small change.

The third is better character consistency across shots. In earlier AI video, a person's face or outfit often drifted between scenes, which made multi-shot stories hard to assemble. "Better" is the report's word, though, not a measured score.

What is still missing

We have no side-by-side samples, so the quality of the longer clips is unverified. Price is not stated, and neither is availability by region. Those details decide whether this is useful to a small studio or only to people with the right account in the right country.

Until Google or independent testers show real output, treat the improvements as promises rather than results.

A fair way to test it

The simplest check is to bring one real brief. Ask for a character who must look identical across three shots, then make two or three rounds of edits. Note how long it takes and what it costs, then compare that with the tool you use now.

Those two tests speak directly to the claims, and they will tell you more than any launch announcement.