Neill Blomkamp has never been shy about chasing the next technological gimmick in service of a genre premise, and his latest project leans into that instinct harder than anything he's made before. This week the District 9 director unveiled a 13-minute sci-fi short loosely based on Peter Watts' 2014 novel, produced through his new AI startup and production company, Barley Studios. Every single shot in the film was generated using ByteDance's video model, with characters whose voices and faces are modeled after real human actors layered on top.
Blomkamp has described the project as a "test start," a proof of concept meant to show what generative AI video is currently capable of, and he's said he wants to eventually tackle a full feature in the same format. Fair enough as an experiment. The problem is that the short is being presented, at least implicitly, as evidence that this format has arrived, and watching it makes the opposite case almost by accident.
The plot follows an American pilot whose helicopter gets shot down during a military conflict between the US and Syria, leaving most of her body catastrophically injured while her head somehow remains intact and functional. That makes her an ideal candidate for a shadowy government program that turns brain-dead soldiers into remote-controlled, weaponized zombies, which is a premise with real potential lurking somewhere inside it, the kind of body-horror military sci-fi that's carried entire franchises before.
What undercuts that potential isn't the writing so much as the texture of the thing itself. Even at its most technically polished, compared with the flood of AI-generated video clips circulating across the internet, the short still carries the specific tells that mark generative video as generative video: a certain smoothed-over quality to motion, faces that read as approximately human rather than specifically human, transitions that feel assembled rather than shot. A real production crew was clearly involved somewhere in the pipeline, coordinating voice performances and assembling a coherent narrative structure around the generated footage, and none of that effort quite manages to paper over the fact that what's on screen doesn't look like footage anyone actually captured.
That gap matters more here than it might in a shorter, more abstract piece of AI-generated content, because a 13-minute military sci-fi short is explicitly trying to earn dread, tension, and a sense of bodily stakes, the exact register where texture and specificity carry the most weight. Slop, as a category, isn't really about production value; it's about a kind of generic, unanchored quality that no amount of polish quite resolves, and by that definition, a well-funded slop production is still slop, just with better lighting.
None of which means the underlying idea, using generative video as a genuine production tool rather than a novelty trick, is dead on arrival. It probably isn't, and there's a reasonable version of this technology a few iterations down the line that closes the gap between "technically impressive" and "actually good." But that version isn't this one, and treating a rough proof of concept as the future of filmmaking, rather than as a rough proof of concept, is exactly the kind of overselling that makes people roll their eyes at AI video generally, whether or not the underlying technology eventually earns a more generous verdict.