Everyone has a story, most never get told because the means are out of reach. An AI-native film studio. From your script or storyboard to finished footage that plays as if it was shot on location.
Written, storyboarded or blocked, all three come out finished. A script, storyboards, a previsualisation, or just an idea. Whichever form your story is in, we pick it up from there and build every shot in service of it. You bring the vision, and we bring the craft, the tools and a career of experience.
Words on a page and a reference file. We settle the look with you, board it, block it and build every shot. Every frame is a decision, and every one of them gets made with you in the room.
Your framing and your cut are already decided. We block those frames in 3D so camera, lens and timing are fixed, then build the finished shots straight from them. Already have a previsualisation? Then nothing is left to interpret.
Every shot built to your camera and your timing, with your approval at each gate. The same shot, carried from the page to the picture.
The street that was torn down, the uniform no one still has, the car that was scrapped in 1991. We rebuild props, locations and relics from the period, checked against archive photos and records, so the reconstruction holds up for anyone who was there. Most of it could not be rebuilt any other way: the place is gone, the object is lost, or building it for real would cost more than the film.
Each shot is framed for both from the start, not cropped at the end. You get a 16:9 master for cinema, pitch meetings and YouTube, and a 9:16 cut for Reels, TikTok and Shorts, from 15 seconds to 3 minutes. Other crops, such as 4:3 or 2.39:1 scope, come from the same shots.
We preserve the physical truth: the right object, the right place, the right conditions. Not a resemblance assembled from memory, but the object as it was, checked against the record before a frame is generated.
Direction costs the same.
Production overhead drops by about 75%.
AI-native tools introduce a new cost structure, and this is the future. Almost the entire budget of a traditional production is overhead: gear, transport, locations, permits, schedule, insurance and the approvals on top. None of it reaches the screen. In an AI-native workflow that overhead falls to a fraction. Compute is a real and recurring cost, but for the first time it no longer dwarfs the work it was meant to serve.
Illustrative proportions, not figures. The direction block is the same size in both columns – that work costs exactly what it always did.
We come from film, and we direct the way film is directed. We are directors who have spent a career making the real thing, and fifteen years of deciding what belongs in a frame goes into every shot. The tools return options and have no opinion about which of them is right. We choose, shot by shot, the way a director on set would.
Every project is treated as its own thing, never a template with the names swapped. Every call gets made in service of your story.
Bring the project you are working on. We’ll turn one minute of it into a finished preview video, free and with no strings attached.
Times shown in your own timezone.
Every frame on this page was built. From a stadium filled to a world invented from nothing, with the same process behind all of it.
Bring the project you are working on. We’ll turn one minute of it into a finished preview video, free and with no strings attached.
Times shown in your own timezone.
We build the footage a story never had.
Nothing is unfilmable anymore. Nothing is out of budget.
AI is our tool. People make every decision.
What does AI-native mean? It means we use AI-tools the way an architect uses a pen or a carpenter uses a hammer. Central to how the work gets made, and with no opinion at all about what it should be. Nobody thinks the pen or the hammer chose where the house goes, how many rooms it has or which way the windows face. A person drew the blueprint. The hammer drove the nails.
Every shot is built from real reference material.
We preserve the physical truth: the right object, the right place, the right conditions. Not a resemblance assembled from memory, but the object as it was, checked against the record before a frame is generated. This is the layer an audience cannot articulate and always notices when it is wrong.
The same world in every shot, held to the same laws. Nothing is re-described shot by shot, so nothing gets the chance to drift, and the world holds across a shot, a scene and an entire story.
The same character in every shot, and the same car, room, uniform and face around them. Each one is built once and carried into every shot that needs it, so shot ninety matches shot one.
Things fall the way things fall. Weight lands, bounces and settles; material reads as material: wet wool, scuffed steel, cold mud. This is the layer an audience cannot articulate and always notices when it is wrong.
How a shot is made: reference in, review, repeat, then edit.
Iteration is the method. Real reference feeds the model. We review what comes back, adjust the reference and the prompt, and generate again. Nothing is accepted because it is close. Then the edit holds continuity across the sequence, which is the part no chat interface gives you.
The tools we use.
More than ten, aimed at one result.
We use more than ten tools to maximise one result. Each does a single job well and none of them decides anything. One tool gives you what that tool makes. Ten aimed at one result give you something none of them makes alone.
What we build the inputs withLeverage comes from knowing which tool to reach for, for which job, at which point in the shot.
The generation is AI-assisted. The edit, the grade and the sound are the same tools every post house uses.
Generic AI output next to our directed result.
Cinematic and specific, not generic and AI-looking. What closes that gap is not a better model. It is ten or more tools aimed at one decision, each doing the one thing it does best.
Straight out of the box – generic results
Directed – specific results
What is slop? The most probable version of itself.
Slop is work optimised for the expected response instead of made from a specific intention. Knowing what works is not the problem, and anyone good at this iterates against what lands. The problem starts when the expected response stops checking the work and starts generating it. Probable, predictable, expected, generic, stereotypical: one thing under five names, and every one of them is made out of what already happened.
The predicted reaction makes the choice. A reaction can only be predicted from reactions that already happened, so the work gets assembled out of things that already worked.
The intention makes the choice. Whether it lands is checked afterwards, as a constraint, never as the thing that decides.
Our typical process. Every project is its own, but the first steps rarely change. This is how it usually runs, from the first call to a working rhythm.
A 30-minute introduction call about what you are making and where it stands.
We return a proof of concept on your project: one finished minute, free and unconditional.
We plug in on top of your existing workflow. Nothing on your side has to be restructured or reorganised.
As we learn how you work, we optimise the workflow together where it helps, which usually means coming in earlier in the process.
Plain and simple: turnarounds of under a week, depending on the length of the project.
Bring the project you are working on. We’ll turn one minute of it into a finished preview video, free and with no strings attached.
Times shown in your own timezone.
Out of the box these models produce the generic.
Aimed at human intent, they make something quite different.
Why most AI video looks the same.
And how direction changes it.
Undirected, AI tools return the average. Directed, they amplify human intent. A model returns the most probable answer. That is not a flaw waiting on the next version, it is the objective working as designed, and more compute makes the answer more average rather than less. Every frame worth putting on screen is an exception to something. Pointing the machinery at one is the whole job.
AI × AI = Imitation
Fed on its own output, every pass moves further toward the middle. The rare and the specific go first.
AI × Human = Exception
Same machinery, different target. A person fixes on a frame the model finds almost impossible, and the iteration is spent getting there.
The value is in the story, not the pixels.
The value of the Mona Lisa is not in the paint. Nothing about that painting is valuable as an object. The pigment cost very little, the panel is poplar, and a perfect print of it is worth next to nothing. What people queue around the block for is not on the surface. It is who made it, when, under what conditions, and what they were reaching for.
A story on screen works the same way. The value is not in the pixels. It is in the story of what it is about, and who decided it should exist.
What a film costs now:
compute, multiplied by direction.
The new commodity is compute. Direction remains as scarce as ever. A badly directed shoot always cost more than a well directed one. Same trucks, same days, more of them wasted. That has not changed. What changed is the thing being multiplied: compute is a commodity priced the same for everyone, so direction is now the whole difference between one quote and another.
Cost = (flights + trucks + set builds) × direction
And behind them: permits, insurance, stage hire, camera packages, per diems, and the approvals stacked on top of the lot. A director who knew the shot got it in three takes. One who did not burned the day, at full rate.
Cost = compute × direction
Everything in the brackets collapses into compute. The multiplier does not collapse into anything. It is the only variable left in the equation, which raises the obvious question of how you measure it.
Shooting ratio: how many tries it takes to get one usable minute.
Directing efficiency has a name: shooting ratio. How much you shoot against how much you use. A documentary might run a hundred to one, a tightly boarded drama five to one. The number never measured the camera, it measured the person deciding, and it is the oldest efficiency metric in the industry. Only the unit changed: feet of film became generations.
Pointed at a single tool, it takes around two hundred generations to get one usable minute, and the result is whatever the tool felt like making.
The leverage comes from knowing what each specialist tool should do, then aiming all of them at one result. Fifteen years of deciding what belongs in a frame is what makes that possible.
We make the film you have in mind.
You make the final call.
We are not trying to make footage that looks good. We are trying to make the film you already had in your head. A handheld documentary and a locked-off period drama are both right, for opposite reasons. Lens, movement, grade, the length of a cut mean nothing on their own, only in relation to the story making the choice.
A model has seen everything ever made and holds no opinion about any of it. It can tell you what is typical. It cannot tell you what is yours. That is why the last call belongs to the author.
Bring the project you are working on. We’ll turn one minute of it into a finished preview video, free and with no strings attached.
Times shown in your own timezone.
01Documentary reconstruction.
What happened, and what could have. Documentary reconstruction and dramatised scenes, both anchored to real places, real periods and real reference. Everything from filling an empty stand to putting a street under weather it never had. What follows is a sample of that, not the whole of it.
02Stylised fiction and new worlds.
New worlds, and the rules that hold them. Science fiction, fantasy, anything set somewhere nobody has been. There is nothing to be faithful to, so consistency becomes the discipline instead: the same place, the same face and the same light in every shot, and a physics the world is never allowed to break.
03Advertising and brand films.
Brand work, built not shot. Your product keeps its exact shape, colour and material. Everything around it is directed: the location, the hour, the weather and the light it sits in.
Four films of our own, built with the same process we use for every client. All of them in active development.
Bring the project you are working on. We’ll turn one minute of it into a finished preview video, free and with no strings attached.
Times shown in your own timezone.
Fabula
Latin: story, tale, narrative.
The oldest word we could find for the thing we are actually selling. Not the tools, not the pipeline — the story, and getting it in front of someone in a form they will sit through.
Directing since 2013 — music videos, commercials and short film, with work for Sony Music and Universal Music. Sets the frame, the intent and the standard every asset is held to.
Assets, research and the client side of every project. Builds the reference boards and the replicas, and runs the approval so nothing is generated before it is agreed.
No account layer, no handoffs, no version of this where you brief someone who then briefs someone else. You talk to whoever is making the thing, which is why notes turn around in hours rather than weeks.
