This is the article an AI film studio probably should not write. We make AI commercials. Spec films, product launches, fashion pieces, made with no brief, just to push the craft. People assume we would say yes to AI video production for everything. We say no about half the time.
That no is worth more than our yes. A studio that tells you AI is the answer to every brief is not advising you. It is selling you its pipeline. The right tool for the wrong job still produces the wrong result, and the bill looks the same either way.
So here is the honest list. The situations where we tell clients to put the AI down and pick up a camera instead. If your project lands in any of these, AI video production is the wrong choice, no matter how good the renders get.
When the product has to be true
This is the big one. If the film’s job is to show your actual product, the thing a customer will hold in their hands, AI is the wrong camera.
AI does not photograph your product. It imagines it. It invents the label, smooths the stitching, straightens the seams, and renders a version of your product that is almost right, which is the most dangerous kind of wrong. Anyone who knows the product spots the lie in a single frame. Everyone else gets a false memory of what you sell, and then the real thing shows up and disappoints them.
This matters most for hero product shots. The macro of the watch face, the pour of the serum, the label catching the light. These are the frames where trust is built or broken. A viewer comparing your ad to your product page and finding a stranger is not a minor glitch. It is a broken promise at the exact moment the customer was ready to believe you.
Our rule is simple: if the film shows the product as the product, shoot the real product. AI can build the world around it, the light, the set, the atmosphere, but the thing being sold has to be the thing. When we build our AI product films, the real packaging is locked in as an unchangeable fact, and the AI directs everything around it. That is the line. Cross it and the film is fiction wearing a logo.
When the story is a person
AI is excellent at objects. It is terrible at people, and not just because of the extra fingers.
The problem is subtler. Human faces carry meaning in movements too small to prompt. A founder talking about the day the company almost died does something with their eyes that no model can invent. A customer describing what your product changed for them pauses in a place a script would never put a pause. Real emotion has a grain to it, an irregularity, and that irregularity is exactly what the audience reads as truth.
Founder stories, testimonials, documentaries, behind-the-brand pieces, any film where a real person is the reason to watch. These do not survive being synthesized. An AI founder is a mascot. An AI testimonial is a lie with good lighting. The audience may not be able to articulate what feels off, but they feel it, and feeling off is the opposite of what a brand film is for.
There is a narrow exception, and it proves the rule. If the people in your film are archetypes, not individuals, silhouettes in a crowd, hands without a face, figures standing in for an idea, AI handles them fine. The moment the film asks you to believe a specific person, that person has to be real. Our brand films live and die on this distinction.
When the process is the proof
Some commercials sell the thing. Some commercials sell the making of the thing.
A craft brand whose film shows the loom, the kiln, the kitchen, the workshop is not decorating. The process is the argument. You are buying the care, the hands, the hours. An AI rendering of craftsmanship is an oxymoron with nice color grading. It shows the idea of care, which is precisely what a skeptical viewer assumes is marketing.
Same with spectacle. A stunt commercial, a live event, a real crowd going wild for a real product launch, these work because they happened. The audience knows a one-take warehouse explosion or a hundred dancers in a street cannot be faked cheaply, and that knowledge is the thrill. AI can render a crowd. It cannot render “this actually happened,” and for some ideas, that is the entire product being sold.
Food is its own special case of this. Food advertising has a whole discipline around real steam, real melt, real pull, because the human appetite has a lie detector that no other sense matches. AI food looks delicious and feels wrong in the same frame. If your commercial’s job is to make someone hungry, use a camera, use a food stylist, use the real thing. Your audience’s stomach will thank you.
When nobody would notice (use AI here)
Fairness demands the other side of the list. There are jobs where AI video production is not just acceptable, it is the better choice.
Worlds that do not exist. A perfume bottle floating through a desert made of glass. A sneaker orbiting a planet. Any concept that lives in imagination rather than documentation is AI’s home turf, and a traditional shoot would spend most of its budget faking what AI renders natively.
Speed and iteration. Need twelve versions of a launch film for twelve markets by Friday? Need to test three concepts before committing to a shoot? AI compresses the expensive part of production, the part where changing your mind costs money. Use it as the sketchbook and the shoot as the painting, or skip the shoot entirely when the sketchbook is enough.
Backgrounds and extensions. The product is real, shot on a table, and everything around it is invented. This is the hybrid sweet spot: real product truth in the foreground, infinite set in the background. Most of the best AI commercial work happening right now lives exactly here.
The pattern is consistent. AI is the right camera when the film sells a feeling, a world, or an idea. A real camera is the right camera when the film sells a fact, a person, or proof. Feelings are renderable. Facts are not.
The Three-Gate Test
Before you brief anything, run the concept through three gates. One yes sends you to a camera.
Gate one: does the product need to be true? If the audience will compare the film to the real thing, a shelf, a site, an unboxing, the product on screen has to be the product. AI imagines. Cameras witness. Choose the witness.
Gate two: is a real person the story? If a specific human being, their face, their voice, their lived experience, is what the audience is meant to believe, that human has to exist. Synthesis reads as synthesis. Nobody bonds with a render.
Gate three: is the process the proof? If the film’s argument is “look how this is made” or “look what really happened,” the making and the happening have to be real. A rendered workshop proves nothing except that someone owns a good GPU.
Three gates, three honest answers. If the concept passes all three with a no, congratulations, AI video production will probably make your film faster, cheaper, and stranger in the best way. If it fails even one, do the honest thing and book the shoot. Your audience will never know what you chose. They will only feel whether it was true.
Talk to us
We make commercials both ways, with cameras and with AI, and our favorite sentence in a client meeting is “you don’t need us for this.” If you bring us a brief, we will tell you honestly which camera it deserves. Talk to us and find out before you spend a dollar.