The most interesting part of AI film isn't the AI

Jonas Ander · 2026-09-15

How we produce branded film, and why the models are the least interesting part of it.

When we started KontentPlus, producing branded images and video at a sensible cost was still genuinely hard, the sort of thing you budgeted for, booked in advance, and used sparingly. Little did we know that both would become integral parts of what we deliver only a year and a half later. I still find that worth pausing on, so I'll say it plainly before getting into the how.

We produce film now. Launch films, product promos, explainers, the short vertical cuts that live and die in a feed. Clients ask how, and they usually expect the answer to be the name of a model. It isn't. The models are the least interesting thing about it.

Here is what actually happens when we make a film.

Story before pixels

Nothing is generated until the idea is settled in words. Before a single frame exists there is a one-page director's sheet: the spine of the story, every scene with its purpose, what the viewer should take from it, how it cuts to the next. You approve it scene by scene, or you send it back. Only then does anything get drawn, and only once the storyboard is approved does anything move.

This sounds like a constraint. It is the opposite. The expensive failure in film has always been generating the wrong thing beautifully. A locked story means the minutes and the money go into shots we already know we want. When a scene isn't working, it goes back to the script, not into another render.

The right tool for every frame

A film is not one material. A face, a room, a slow push across water want a generative model. A price, a headline, a logo, a caption, a chart, a product screen must be exact, so we never let a model guess at them. They are composited on top as real graphics: pixel-precise, in the real typeface, with the real logo.

The rule we work to is plain. If a shot doesn't need photographed-looking content, no video model is opened at all. Type, motion graphics and data are cheaper, faster, perfectly repeatable, and they never spell the brand name wrong.

The stack is a commodity, the system is the point

We use different models for different jobs: one for stills, another for motion, another again for the harder shots, a spoken line with correct lip-sync, or two camera angles inside a single continuous take. That last one was impossible a year ago and will be ordinary a year from now.

Which is exactly why we don't build on any of them. We build around them. Every model we use will be replaced by a better one within weeks, at the same moment, for everyone. What lasts, and what compounds, is the operation around the models: the brand rules, the approved voice, the shot discipline, the memory of every note you gave on the last film. That is the asset. The model is a rental.

One build, many cuts

The finished film is assembled as a composition, not baked into a single file. Timing, text and transitions are set in the edit, frame by frame, at no extra cost. We don't re-generate an expensive clip to fix a beat that lands half a second late.

Because it's a build and not a bake, everything downstream is nearly free. The 16:9 master yields the 9:16 vertical, the six-second bumper, the silent captioned version and the Swedish narration from the same source, without touching the footage. One film becomes a small family of films.

What it actually takes

A finished launch film, reference, script, storyboard, generated footage, voice, music, a graded 1080p master and a vertical cut, is roughly a working day, with your approvals built into it, at a generation cost closer to a good lunch than a production budget. The real saving isn't the money. It is that the whole thing stays legible: at every step you can see what it will cost before you spend it, and change it while it's still cheap to change.

Why this sits with us

None of this is a side venture. Film, for us, is content operations with a camera on it: the same instinct that runs through everything we build, a clear system underneath so your judgement goes into the part that deserves it. The frontier is genuinely moving fast. The advantage was never being first to a new model. It was having somewhere disciplined to put it.

If you have a film in mind, tell us the one thing it should make a viewer feel. We'll show you the storyboard before anything moves.