What a creative technologist does on an AI production
Some shots cannot be made by any of the usual routes: a shoot does not fit the schedule, 3D is too slow, and AI alone will not stay consistent. This explains who solves that, when to bring them in, and how it worked on a national TV commercial.
A creative technologist is the person on a production who works out how a shot can be made when no single tool can make it. The role joins a filmmaker's judgment with AI, 3D, compositing and custom scripts, and builds the route between them. On small jobs it often sits with the director.
01 · The Role
What the role does
Creative technology can sound abstract when it is described as a job title. In practice it starts when a brief arrives with a production problem that the standard route cannot solve. The deadline is fixed and so is the client's expectation. You cannot simply say that the tool does not exist yet. You have to work out what the problem is.
Then you decide what each available tool can do well, what it cannot do, and where the gap sits between them. Sometimes the answer is a shoot. Sometimes it is 3D. Sometimes it is AI, compositing, code, automation, or a new process built specifically for that project.
The last part is the one that separates the role from a tool operator. When the route does not exist, a creative technologist builds it: a workflow, a script, a way of passing work from one program to the next so that an approved decision stays approved.
02 · The Timing
When to bring one in
The usual sign is a shot that every standard route fails. A shoot does not fit the schedule, a full 3D build is too slow, and AI alone cannot hold the consistency the client needs.
The other sign is AI work that looks right in one frame and falls apart across a sequence or once the notes come in. In both cases someone has to diagnose the problem and decide whether the work needs another generation, a control layer, 3D, compositing, a better workflow, or a different production route entirely.
It is easier when that person is in the room before the approach has been promised to anyone, because the route decides what can still be changed later.
03 · A Worked Example
A worked example: the hundred-person crowd that did not need a shoot day
A crowd like this normally means a full shoot day: around a hundred background artists, a crew, a hangar. This production had no shoot day to give it. So the crowd was built as a system.
The example comes from the Gigaclear commercial. The whole production ran for about a month, and this was one hard problem inside it: put one hundred distinct, branded people inside a hangar, seen from multiple camera angles.
What a hundred-person crowd takes
To make that crowd for real, you find and cast one hundred people, get them to the location, and put them in wardrobe with the client logo applied. You organise transport, releases, security, makeup and catering. You hire the hangar, a full crew, and the equipment to light and shoot the space.
Then you direct the crowd, capture every required angle, and hope the day gives you enough usable material. It is a full shoot day with its own line in the budget before the shot has even entered post-production.
The schedule had about a day for it.
3D gave control and took too long
Conventional 3D would have given us control. We could build and texture people, create clothing variation, place logos, arrange the crowd, and render it from any angle.
But one hundred individual people are not a quick 3D task, even with existing character assets. The crowd cannot simply be cloned if the camera sees enough of it. The faces, body types, clothing, logos, rows and camera angles all have to hold together. 3D could meet the consistency requirement. It could not meet the deadline.
AI alone was also the wrong answer
Then I experimented with AI. It can generate a crowd quickly. Ask for a group of people inside a hangar and it will give you one. Ask for another version and it will give you a different one. The faces, the outfits and the number of people all change, and the logos become almost-logos. Change the camera angle and the entire crowd is reinvented.
That is fine for a single image. For a commercial, the crowd needed to survive the next shot and client approval, and to stay useful when the camera moved. The logo had to be correct because it belonged to the client. The people had to look different, but they also had to belong to the same world.
A physical shoot needed a day of its own that the production did not have. A 3D build was too slow. AI alone could not hold the crowd consistent.
The answer had to be built
First, I worked out what had to remain fixed. The crowd needed an approved visual language: how they sat, what they wore, the range of faces and body types, and exactly where the client branding had to sit.
Then I built a workflow that generated crowd rows rather than isolated people. Instead of creating one character at a time, I could specify the number of rows I needed and the kind of variation required: different people, different builds, different appearances, and the same visual rules.
That gave me the crowd as front and back image material.
Then came the logo. AI can produce something that resembles a logo, but resemblance is not enough when the logo belongs to the company paying for the commercial. So I built another workflow to place the approved logo in the exact required position across the crowd material.
At that point, the crowd had the right people, the right wardrobe, and the right branding. But it was still only image material. The hangar needed three-dimensional assets.
The right amount of detail
The front and back crowd images were converted into usable 3D assets.
The important decision here was not to build more detail than the shot needed. The crowd was not going to be examined as one hundred hero characters in close-up. It needed to hold up in the composition, from the camera angles used, at the distance the audience would see it.
From there the crowd moved from approved images into textured 3D forms. Now it could live inside the hangar, take light, hold its place in the composition, and be rendered from the required angles.
The last problem was layout. Placing rows of crowd assets manually inside Cinema 4D would have created another bottleneck. We needed a clear centre line for another character to move through. We needed enough depth for the hangar to feel full. We needed spacing that looked intentional rather than procedural.
So I built the final bridge: a script that connected a local model to Cinema 4D through an MCP workflow. I could tell it how many rows were needed, how they should be arranged, and where the gaps had to be. It laid out the crowd inside the scene.
There was still adjustment afterward, because there is always adjustment afterward. But the hard part was over.
04 · The Value
Scale became a parameter
The value was that the problem stopped being a one-off.
If the agency needed more people, the system could make more people. If the layout changed, the system could be adjusted. If the crowd needed a different visual mix, the workflow already contained the structure needed to create it.
One hundred people had been the bottleneck. After the system existed, the size of the crowd was a setting that could be changed.
A physical crowd shoot gives you one day of footage. A conventional CGI build gives you control, but at a much slower pace. This system matched the project: it held up through approval and kept working when the brief changed.
The crowd worked because each part of the job was given to the right method. AI generated variation quickly. A controlled workflow made that variation usable. A logo workflow protected the brand. Image-to-3D conversion created assets for the scene. Cinema 4D gave those assets position, camera logic, and control. The script removed the repetitive work of placing every row by hand.
05 · What You Get
One person who decides what is generated, filmed or built
I came into film production in 2013. Since then, I have worked across editing, motion design, VFX, 3D, compositing, creative direction, and AI production. There is more on my background on the profile page.
For you, that range means one person decides what is generated, what is filmed and what is built, and then makes it. The decision and the work do not pass between departments, so nothing is lost in a handoff.
It also helps when the notes come in. The person who built the route already knows what a change to one shot will touch in the others, which is the subject of The One Small Change.
Sometimes the route already exists. When it does not, the work is to build it between the brief and the finished shot. If you have a job like that, these are the ways to work together.
Frequently Asked Questions
What is a creative technologist?
A creative technologist combines creative judgment, production experience, and technical systems to solve a problem that no single existing tool can solve alone.
When should you bring in a creative technologist?
When a shot fails every standard route, or when AI work looks right in one frame and falls apart across a sequence or once the notes come in. It is easier before the approach has been promised to anyone, because the route decides what can still be changed later.
Is a creative technologist a programmer?
Not necessarily. Scripts, APIs and automation can be part of the toolkit, but the role begins with production judgment: understanding what the brief needs and building the right route to deliver it.
What is the difference between a creative technologist and a creative director?
A creative director decides what the work should say and feel. A creative technologist works out how it can be made reliably, using the right combination of production, AI, 3D, VFX and custom systems.
Who fixes AI video when it is inconsistent or not production-ready?
Someone who can diagnose the actual problem: whether the work needs another generation, a control layer, 3D, compositing, a better workflow, or a different production route entirely.
Do creative technologists build their own tools?
When the project needs them. The crowd system in this article connected AI generation, logo control, image-to-3D conversion and Cinema 4D layout because no off-the-shelf tool could solve the brief alone.
Send the script or the deck. You get a reply within a day: what is realistic and how I would make it. This first read is free.
Send the brief → See how the Gigaclear commercial was made →