Cully Hill Boys, a feature-length AI film uploaded online in mid-August 2026 by the AI platform Higgsfield, looks almost like a real movie. Its three-act story of Londoner daydreamers feels familiar, and its kinetic, stylish shots are reminiscent of directors like Edgar Wright or Guy Ritchie. According to the source material that created it, however, its best qualities are direct imports from humans.
According to reporting by The Verge, the script was written by human screenwriter Tim Planagan and landed on the Black List. The prompts that generated the video frequently name-check Wright and Ritchie. And while AI tools did the rendering, a team of nine directors led by Adilet Abish and Aitore Zholdaskali curated and assembled the output. The final film, while a leap forward for AI video generation, serves as a massive public prompt study to showcase Higgsfield's tools, revealing both the staggering potential and the defining limitations of AI filmmaking today.
The Fragile Illusion of AI Narrative Coherence
For an AI, a movie script is a complex puzzle of cause and effect, character continuity, and emotional logic far beyond stitching together cool-looking shots. The AI models powering projects like Cully Hill Boys excel at generating individual, impressive 30-second clips. But maintaining a character's appearance, clothing, and personality across dozens of shots stumps them.
The Higgsfield team's workflow, as described in the source, is telling. They had to split their work across different AI models, "because the rules of one job poison the other." One instance of Anthropic’s Claude was dedicated solely to crafting prompts for images, another for video clips. This is not the unified vision of a director but the work of a technical assembly line managing disjointed fiefdoms of specialized bots. The plot may move, but it does so because a human script provided the map, not because the AI grasps cause and effect.
As we saw in earlier coverage of AI Apps Automating 30% of Your Workweek for You, these tools are powerful but rigid automations of discrete tasks. They do not understand the interconnected whole.
"It definitely feels like Cully Hill Boys is a stronger film than Hell Grind because its core narrative was crafted by an experienced, human storyteller with an understanding of proper pacing," writes The Verge's Charles Pulliam-Moore.
Result: You get scenes that technically flow, but characters who lack genuine chemistry, and bizarre details like "incomprehensible text in books" that break the illusion.
The Human Curator Is Still the Director
Cully Hill Boys is less a film made by AI and more a film assembled from AI outputs. The critical creative work happened before and after the models ran.
The Creative Inputs Were Human:
- Story: A traditional, human-written screenplay.
- Vision: Prompts explicitly directing the AI to mimic the styles of specific, beloved human directors ("GUY RITCHIE COVERAGE", "DYNAMIC clip-style (Edgar Wright snap.)").
- "Actors": Likenesses licensed from real people, streamer Mikyle “N3on” Rafiq, creator Matt Kiatipis, and UFC fighter Israel Adesanya.
The Critical Curation Was Human:
The AI didn't "direct" the zippy fantasy sequences or decide which smash cut worked best. Higgsfield published many of the prompts, revealing they iterated on individual shots dozens of times. A human, or a team of them, then selected the "best" take. This is curation, not creation in the directorial sense. It's akin to a photographer sorting through thousands of auto-generated images to find the one perfect shot.
This process was also evident in Higgsfield's earlier film Hell Grind, which Variety described as having "shockingly realistic" visuals but a "nonsensical" plot. Its most memorable flaw, as noted by TechRadar, was AI-generated dialogue that shattered immersion, a character facing imminent demise inexplicably exclaiming "Cool!"
The Uncreatable Nuance: What AI Still Can't Do
The cracks in Cully Hill Boys expose the specific creative tasks that remain firmly in the human domain.
Character Consistency & Emotion: AI can generate a face that looks sad based on a prompt, but it cannot build a character's emotional arc across 90 minutes. The subtle facial tics, the unspoken chemistry between actors, the lived-in weariness of a person, these are products of lived experience and performance, not statistical prediction.
Understanding "Why" and "How": Current generative models are brilliant pattern matchers. They are trained on massive datasets of film end products. But they do not understand the underlying mechanics of why a scene works. They can't grasp comedic timing, motivation, or subtext. A prompt can ask for "a tense argument," but the AI cannot layer that argument with years of fictional backstory.
Original Creative Spark: The most crucial human part is the very beginning: the original idea. The concept of three East London daydreamers using fantasy as an escape came from a human mind. The AI's job was merely to illustrate that idea, often imperfectly. It is a powerful but unoriginal illustrator.
This distinction is critical when evaluating new tools, just as developers weigh Should You Use VS Code or Visual Studio in 2026?, the core work defines the tool's role, not the other way around.
AI's Role: The Hyper-Powered Pre-Visualization Tool
Cully Hill Boys is a compelling proof of concept, but its true value isn't as a standalone film. It's a demonstration of AI's emerging role as a hyper-powered pre-visualization and rapid-prototyping tool for human filmmakers.
Consider this potential workflow:
| Traditional Filmmaking Step | AI-Augmented Alternative |
|---|---|
| Storyboard artists draw keyframes for weeks. | A director uses prompts like "Edgar Wright-style snap zoom on a nervous man in a pub" and gets multiple, fully animated video versions in minutes. |
| A director describes a complex stunt sequence to a team. | The director generates a rough, 30-second AI video of the sequence to communicate the vision clearly to the stunt and VFX teams. |
| Waiting for location scouting and set builds to see a scene's feel. | AI generates the characters into a mocked-up version of the intended location, allowing for immediate pacing and blocking tests. |
This is the real collaboration pattern emerging. The current state of AI video is not a replacement for cinematographers, screenwriters, or actors. It is a new, incredibly fast tool for exploring visual ideas, one that still requires a human artist to have an idea worth exploring in the first place.
The output will only be as compelling as the human direction behind it. The most promising path forward doesn't pit AI against human creativity but positions AI as a new, potent medium through which that creativity can be expressed and iterated upon at unprecedented speed. The "best parts" will, for the foreseeable future, remain human.
Why This Changes Everything
- Demonstrates AI's current role as a visual tool, not a creative director, reshaping film production workflows.
- Highlights that human curation and assembly remain critical for narrative coherence, challenging predictions of fully autonomous AI filmmaking.
- Reveals AI video generation as a 'massive public prompt study' that exposes both potential and limitations for the entertainment industry.
Originally published on XOOMAR. For more news and analysis, visit XOOMAR.
Top comments (0)