Join the waitlist and win $25 credit! Apply now
Manifesto
Why we built this
The problem we set out to solve predates artificial intelligence by roughly a century: the distance between the film a director can envision and the film a production can afford to realize.
Scenes are lost in that distance. It closes in budget meetings, in scheduling conflicts, and in the daily calculation of which setup is sacrificed so the schedule survives. When a scene is cut, the day required to produce it is cut with it — the crew, the cast, the hours, the work.
That distance is not a fact of nature. It is a tooling failure. Film production runs on instruments that have not kept pace with what the work demands, and every year the gap widens. ORB was built to close it.
Where we stand on generative tools
A promise is being marketed to this industry: that filmmaking can be substituted by description. That a sentence of text can stand in for a set, a department, a discipline, and — not incidentally — for everyone who would need to be paid.
The promise is wrong on both counts, and we will say so plainly.
It is wrong about the work. A film is not a description of a film. The distinction between a shot that succeeds and one that does not resides in a lens selection, a stop of light, an eyeline, a decision about where the camera stops moving. These are not adjectives. They are judgments, made by professionals who spent years learning to make them.
It is wrong about the images. When a text-prompt system generates a performer, that performer is a composite of people who did not consent, were not paid, and will not be credited. That is not a technical footnote. It is the industry's central objection, and the industry is correct.
We are not asking anyone to set aside their skepticism toward AI. It was earned. We built our system on the opposite premise, and the architecture shows it.
What we built
ORB is set-based, not prompt-based.
That distinction is architectural. Within ORB, a scene is not a paragraph. It is a coordinate space: real geometry, real camera mathematics, real focal lengths, real distances, real blocking. The filmmaker positions the camera and selects the lens. The system does not interpret intent, because nothing was described. The scene was staged.
Nothing in ORB originates from nowhere. Through the ORB Scan application, productions capture their own locations and their own performers. Not an approximation of a room — the room. Not a likeness averaged from strangers — the performer who was cast, who signed, who is on the call sheet, who is paid and credited.
That decision is what makes the platform work. It is also what makes it defensible, and we chose it before anyone asked us to.
The platform runs three workflows, which are one principle at three scales.
Previsualization. Scan the location, block the scene, plan the coverage. Day one begins with the knowledge that the shot works. Fewer redundant setups, fewer overtime hours, fewer decisions made in front of a waiting crew.
Reshoot. Convert an existing shot into a digital twin and rebuild it with revised intent — a different angle, a different lens, a different lighting design. The source is the production's own footage. This recovers the scene that would otherwise ship at seventy percent, or be deferred to a reshoot day no schedule can absorb.
Fully synthetic production. Scanned performers, scanned environments, synthetic camera and light, for projects that would not have been financed at all.
One pipeline. One geometry. One chain of consent.
We build with directors, by design
ORB is developed in continuing collaboration with working directors from studio and independent production. This is not an endorsement arrangement — it is the method.
Cinematographers tell us where the camera model is wrong, and we correct it. Producers separate the parts of the workflow that are presentation from the parts that recover a day, and we cut the former. Directors tell us when an output reads as a film and when it reads as software.
Software people cannot originate a tool for this industry alone, and we have never pretended otherwise. That is not a limitation we apologize for; it is a decision about how the product gets built. Every serious instrument in film history was shaped by the people operating it, and ORB is being built the same way.
We work with filmmakers who bring us hard constraints. If that describes you, the door is open.
On fully synthetic production
We are not going to file this workflow quietly under a feature list. Here is our position.
The entry point into this industry has closed. There was once a progression — a short on borrowed equipment, then a music video, then a commercial, then a feature. It is largely gone. For a filmmaker without capital, the alternative to a synthetic production is not a smaller crew. It is no production.
We built this workflow for that filmmaker. It is not built to displace a crew on a financed production. That work exists, it is held by people who do it well, and we are not competing for it. We built it to put the first step back.
A meaningful proportion of the people who complete their first film in ORB will be hiring crews within a few years. That is not a qualification. That is the point of building it.
What we commit to
These are enforceable positions, not sentiments.
A performer's scan belongs to the performer. Created with explicit, specific consent. Withdrawable. Deletable.
One scan, one production. Use on a subsequent project requires new consent and new compensation. No exceptions.
No scan — performer or location — trains our models. Ever.
Location scans and set data are the production's property, not ours.
We do not train on footage acquired without permission.
We do not build digital resurrection. Not for a studio, not for an estate.
We will never market ORB on the promise of eliminating a department.
We set these before we had customers asking for them, and they are not negotiable at any deal size.
