AIで制作された映画を内側から検証——最も優れた部分はすべて人間の手によるものだったI looked inside an AI generated movie, and the best parts were all human

Generative artificial intelligence has advanced rapidly from producing still images to assembling moving pictures, and a new short film built with tools from the startup Higgsfield offers a useful test case for how far that shift has actually gone. A behind-the-scenes examination of the project reveals that its most convincing moments appear to owe more to the humans involved—writers, performers, and a director—than to the software that generated the footage.

The film, "Boys Black List," follows a trio of bumbling English lads who fantasize about becoming megastars while knocking back a few pints in a grimy London pub. As they chortle and try to one-up one another, each conjures an increasingly glitzy vision of future stardom. That premise, and the comic timing that sells it, was shaped by people rather than by prompts, and it is the part of the film that lands most reliably.

Higgsfield is one of a growing crop of companies building generative video tools aimed at filmmakers and marketers. Its platform, like rival systems, turns text descriptions and reference images into short animated clips, offering controls over camera movement and visual style. The appeal is obvious: a small team, or even a single creator, can in principle generate scenes that once required a crew, a set, and a budget. In practice, the technology still struggles with the things audiences notice most.

According to the examination, the strongest elements of "Boys Black List"—its story, its characters, and its direction—came from human creators, while the AI handled the generation of imagery. That division of labor exposes the current limits of the tools. Maintaining a consistent face, wardrobe, or setting across multiple shots remains difficult, and the small distortions that generative models introduce can undercut a performance or break the illusion of a scene. Human judgment was needed to script the jokes, guide the acting, decide what to keep, and stitch the results into something coherent.

The finding echoes a broader pattern across the industry. Text-to-video systems such as OpenAI's Sora, Runway's Gen models, Google's Veo, and Luma's Dream Machine have all produced striking clips in isolation, but sustaining narrative, emotion, and continuity across a full scene is a far harder problem. Many working filmmakers have described these tools as accelerators for specific tasks—concept art, backgrounds, transitions—rather than replacements for the creative decisions that give a film its shape.

The context matters because expectations around AI filmmaking have run high. Studios, advertisers, and independent creators are all experimenting with generative video, and the debate over its role has been sharpened by labor concerns, including the scrutiny that unions such as SAG-AFTRA and the Writers Guild of America have applied to how AI is used in production. A finished project like "Boys Black List" is valuable precisely because it tests marketing claims against an actual piece of work rather than a curated demo reel.

None of this suggests the technology is standing still. The pace of improvement in generative video over the past two years has been steep, and features that seemed out of reach recently—longer clips, smoother motion, tighter control—are appearing quickly. But the examination of Higgsfield's film points to a distinction that is likely to persist for some time: generative models can produce arresting images, yet the choices about what a story should say, how a character should feel, and when a joke should land still rest with people.

For creators weighing whether to adopt these tools, the takeaway appears to be pragmatic rather than dismissive. AI can lower the cost of getting an idea onto the screen, but it does not supply the idea, and on the evidence of this film it does not replace the human craft that makes the best moments work.