AI video has spent quite a while being impressive for five seconds and then slightly terrifying when you look too closely. It is getting more useful.
The trick we keep coming back to is starting with something real. Give the model a strong image, describe one short scene and let it create movement around that rather than asking it to invent an entire film from scratch.
Tools such as Veo make it possible to turn still imagery into short clips that are useful for ideas, social content, storyboards and visual experiments. You still watch hands, faces, logos and product details very carefully.
But it has crossed the point where the interesting question is less “can AI make video?” and more “where can a 5 or 10 second clip save us from having to shoot something?”