A human director shapes a film made with generative tools by deciding what the film is for, what counts as a good take, and what never reaches the audience. The tools produce candidates quickly. They don't choose between them. If you're commissioning an AI-native film, the useful question is not "how much did the software do?" but "who decided, at each stage, that this was right?"
This piece sets out where that judgment sits, what a serious sceptic would say against it, and what it means for the brief you hand a studio.
Where does a director still decide?
Four places, in order of how much they affect the finished film.
The premise and the boundaries. Before any frame is generated, someone has to settle the audience, the emotion the film should leave behind and the lines it must not cross: claims, cultural cues, competitor territory. A generative model will happily produce a polished scene that's wrong for the brand. Only a person who understands the brief can say so.
The reference pack. Generative work is steered by what it's shown: look references, wardrobe, location mood, a character description. A director builds and approves that pack. Two studios using the same tools can produce very different films because their reference packs differ.
Performance and pacing. Where a character's glance lands, how long a shot holds, when the music drops out. These choices are what make a film feel directed rather than assembled. A model can offer ten versions of a moment. A director knows which one carries the scene.
The approved cut. Selection is authorship. Across dozens of generated takes, the director rejects most of them, and the film is defined as much by what is thrown away as by what is kept.
What is the strongest argument against this view?
The sceptic says that when the tool generates the image, performance and camera move from a text description, the director is closer to a client reviewing options than to a filmmaker. That's a fair challenge, and it holds in one case: when "direction" means writing a single prompt and accepting the first result.
It doesn't hold when the process has stages. A director who writes the script, builds the reference pack, rejects takes against a shot list, corrects continuity and supervises the edit and sound is making the same category of decisions a director makes on a physical set, with a different mechanism for producing the raw material. The honest position is narrower than "AI has no effect on direction". Direction has moved from commanding a crew on a day of shooting to curating and correcting a much larger volume of options, and the skill of rejecting well matters more than before.
There is also a legal dimension that makes human decision-making more than a creative preference. In its January 2025 report on copyrightability, the US Copyright Office concluded that, with current technology, prompts alone don't give a user enough control to be the author of the output, according to a Copyright Alliance summary of the report. The same summary notes that human-authored expressive inputs and human modification or arrangement of outputs can support protection. That's US guidance, not Indian or Gulf law, and it's not legal advice. But it points the same way as good craft: a film shaped by documented human choices is easier to defend than one that was simply generated.
What does "human-directed" look like in practice?
A useful test is whether you can see the director's hand in the paper trail. Ask a studio for:
- A written script and a shot list that existed before generation began
- The reference pack the team worked from, and who approved it
- Review rounds with notes: what was rejected and why
- The person named as responsible for continuity across scenes
- The final approved cut, signed off before delivery
If a studio can't show these, "human-directed" is a slogan. If it can, you have evidence that decisions were made deliberately, which also helps when your legal or brand team asks how the film was produced.
What changes for the brand team commissioning the film?
Your role shifts earlier. On a traditional shoot, much of the creative risk is locked in at pre-production and then physically constrained by the shoot day. With generative tools, revisions are cheaper in time, so there is a temptation to leave decisions open. That's a mistake. A film with no locked premise drifts, and review rounds multiply.
Practical habits that help:
1. Bring a brief with a single objective. One audience, one emotion, one action you want afterwards.
2. Approve the reference pack in writing before the team generates at volume.
3. Fix the number of review rounds and agree who has final say on your side.
4. Review on the device your audience uses. A vertical film for mobile reads differently on a desktop monitor.
5. Check for artifacts at full speed and frame by frame. Hands, text on signage and background faces are common trouble spots, so ask the team how they check them.
Our own work spans AI brand films, AI micro drama production, AI native / OTT production and performance marketing ads. Across those formats we have created 50+ brand films and ads and produced 10+ episodes and 200hr+ of content for OTT, and the direction question comes up on every one of them. The answer is always the same: a named person owns the creative call.
FAQ
No. Generative tools replace some of the labour of producing raw footage, not the decisions about story, performance, pacing and what to keep. Those decisions still need a person who understands the brief and can be held accountable for the result.
Ask for the pre-generation script and shot list, the approved reference pack, the review notes and the named person who signed off the final cut. A studio that works properly will have these to hand.
It depends on the country and on how much human authorship went into it. The US Copyright Office has said prompts alone are not enough, per the Copyright Alliance summary above. Take advice from a lawyer in your market before relying on ownership of a generated film.
The decisions are smaller but the principle is the same. Someone still has to choose the hook, the pacing and which of many variants represents the brand. The difference is volume: more variants, faster review.
A script or rough storyline, two or three reference films you admire, the language versions you need, the platform where the film will run and one measurable objective. That gives a director something to direct. If you have a script or reference film in hand, schedule a call and we'll talk through how we would direct it.
