A new version of a video generation model began producing about ten seconds of video from a single image, reported in the second week of August.
Ten seconds sounds small and is the unit that matters. It's the typical length of a clip used in a social post, and long enough to show a complete action.
The practical gain for anyone producing content is time. Material that required filming, a crew and editing can now be produced from an image, which changes the cost of testing an idea before investing in it.
The problem growing alongside it is verification, and it already appeared concretely in the period, when a tool allowing AI-altered satellite imagery was pulled a day after launch for turning a trusted source into a potential misinformation instrument.
The difference between the two cases is the context of use. Generated video for an ad is production; generated video circulating as a record of an event is something else, and no watermark helps once the screenshot has been taken.