From Novelty to Art Form: A Director’s Guide to AI Filmmaking

Why AI Filmmaking Still Demands a Director's Hand

That distinction is where Scott Palmer, Creative Director at Cramer, starts every conversation about AI video generation — and it’s the distinction most of the current hype in this space glosses over entirely. Scott has spent much of his career directing shoots and building video concepts for clients. Over the last two years, he has dedicated himself to understanding where AI video generation can genuinely integrate into the craft of filmmaking and where it can’t — testing platforms, documenting limits, and producing work that has cleared the most unforgiving quality threshold in live events: a widescreen conference stage in front of thousands of people.

His take on where the industry is headed is measured, specific, and worth paying attention to precisely because it resists the hype in both directions.

75% Is Not the Finish Line

Scott is direct about where AI video generation currently lives: it takes you most of the way there. The photo-realistic environments, the establishing shots, the atmospheric footage that makes a story feel grounded in a real world — the tools handle that with a speed and flexibility that traditional production can’t match.

Consider what it takes to produce a short film requiring three distinct international settings — airports, city streets, hotel interiors across multiple continents — each demanding authentic environments and the kind of visual specificity that traditionally puts a location scout on a plane. Captured traditionally, a project at that scope means a full production crew, months of pre-production logistics, location permits across multiple countries, and a travel budget that rivals the creative budget. Built with AI-assisted generation and finished by hand where the tools fell short, the same creative scope was recently delivered by a two-person directorial team on a tight conference deadline. No flights. No permits. No months of pre-production. The locations existed in the world — the production didn’t have to go to them. That’s the capability shift. Work that wasn’t viable at that budget and timeline now is.

That last 25%, though, is where the craft lives — and it demands a filmmaker’s eye. Crowd scenes default toward uniformity unless directed specifically for diversity in age, body type, and wardrobe. Scale relationships between objects break down in ways that read as immediately wrong to anyone trained to notice. Custom aspect ratios — the widescreen panoramic formats that dominate large-scale event production — fall outside what most platforms output cleanly, pushing sequences back into traditional post-production regardless of how the rest of the piece was built. File management compounds all of it: iteration is fast, volume is high, and tracking which version of which generated clip is the right one requires organizational discipline the tools themselves don’t supply.

Understanding those limits before you promise a client something the pipeline can’t deliver isn’t a criticism of the technology. It’s the condition for using it well.

"We have to move past the novelty and toward the art form."

The Storyboard Doesn’t Go Away. It Becomes More Important.

There’s a version of the AI filmmaking conversation that treats storyboarding as a legacy step — something the tools are eventually going to absorb. Scott pushes back on this directly. If anything, the storyboard matters more when AI is doing the generating, not less.

AI video generation is extraordinarily capable at producing individual moments — a shot of a city street at dusk, an interior space with a specific quality of light, an atmospheric sequence that feels lived-in and real. What it cannot do is decide what those moments mean in sequence, determine the throughline of a piece, or figure out why one image should follow another. That work belongs to the filmmaker. The storyboard is where it gets done.

"Without a storyboard, you're just generating. You might generate something beautiful. But you won't generate something that means anything — because you never decided what it was supposed to mean."

The Idea Is Still the Job

What AI has not changed is where the real work begins. The tools generate footage. They do not generate concepts, emotional logic, or the specific decision about what a piece is trying to make an audience feel and why. Those remain the filmmaker’s territory — and the quality of the output is almost entirely determined by the quality of the thinking that precedes the first prompt.

The filmmaking vocabulary that has always mattered applies directly to how these tools are used. Composition, pacing, the logic of what the camera chooses to show and when — understanding that language is what produces results that feel directed rather than generated. Scott draws on the same instincts he’d bring to a traditional shoot: what does this scene need to do, what is the audience supposed to understand when it’s over, and what visual decision serves that goal. The platform is different. The questions are identical.

The wide availability of AI video tools has accelerated the technology in amazing ways — but the idea of “prompt engineers” replacing actual creative teams is ridiculous. The best AI video results will always come from real human creatives – filmmakers, animators, and sound designers.

That’s a harder standard than most current AI filmmaking conversation is willing to hold itself to. The question is never “can we generate this?” It’s always “does it serve the story?” For filmmakers willing to work at that level — to treat the storyboard as the foundation, the human craft as the finish line, and the idea as the thing no platform can replace — the tools are genuinely expanding what’s possible. For everyone else, the novelty wears off faster than the render time.

Have a project in mind?

Let's chat