How we make
what you can't film.
This is how we work, one workflow at a time. Each method here has been tested on our own films and brought into productions alongside the crew, the edit and the people who sign it off.
Shortlisted · Best Use of AI (Workflow) · Broadcast Tech Innovation Awards 2026Want the finished films? See our work →
The floor is moving.
Here's where we stand.
British television is shifting under our feet: audiences, budgets, crews. The new tools are already on screen, from Hollywood to Britain. This film follows one production from script to screen through every workflow on this page. The production is fictional. The tools are not.
Where We Stand · 4:43 · Endurance is a fictional production · everyone on screen is AI-generated · narration: AI voice · sources in the film's credits
Use the channel guide, or focus the TV and use left/right arrows. The whole film opens full screen.
Television keeps
changing channels.
Live broadcast, the living room, colour, the edit. Every few decades the way television is made changes, and the people who adopt it first shape what comes next. Scroll through ninety years of it, made as one continuous camera move.
All images AI-generated · a studio film by Channel Changers
Shoot it in a grey room,
finish it anywhere.
Period streets, lost buildings, places you'd never get a permit for.
The actor performs for real on a simple grey set. We build the finished world on the first frame, get it signed off as a still, then carry the whole take into that world with the performance and the camera move intact.
It suits period drama, historical reconstruction and any location that's gone, too costly or out of reach. The example here is The Watcher, a greybox test for a Victorian reconstruction.
Workflow demo · all images AI-generated
The workflow in one minute, narrated by Sam Wilkinson.
Stills in GPT Image · video in Omni 1.1 Flash on Google Flow · the transform in Omni V2VLive Grey: three shows, one grey stage
Our six-minute film takes the grey room live. A phone on top of the camera tracks where it is about sixty times a second, so the director can watch a rough version of the finished world on a laptop while the take is still running. Then the stage changes channel three times: a magician in a music hall, a planning officer measuring a dragon, and a vertical drama in a lift that opens onto a new world on every floor.
Channel Changers film · everyone on screen is AI-generated
Live Grey · 5 min 55 sec. The kit, the tracking, two ways to finish a shot, and the small print.
Sony FX6 for picture · iPhone tracking and LiDAR · finished in Omni, Seedance and RemotionHow a live grey shoot works
Virtual production used to mean building the set, travelling to the location, or hiring an LED volume built for blockbusters. Live Grey keeps the performance real and moves the world into post, but it gives the director a look at that world on the day, so nobody waits weeks to find out whether a take works.
- 01
The kit
A Sony FX6 records the picture. An iPhone mounted on top doesn't film; it logs where the camera is, about sixty times a second. Calibrate it once per lens and wherever the phone goes, the camera goes.
- 02
One clock
A Tentacle timecode box gives the camera and the phone the same clock, so every frame of picture lines up with a camera position. The stage, the camera view, the phone's view and a rough finished world all run in sync.
- 03
Judge it live
The rough world streams to any laptop or tablet on the same Wi-Fi. The director watches the performance inside the finished space as it happens, and notes go straight onto the slate: "Bow lands late. Go again."
- 04
Crosses on the walls
Point the phone at a bare grey wall and it loses its grip, so the world slides. Simple crosses on every wall the camera will see hold the track, even when the camera turns right round.
- 05
Light for the world
The stage is lit flat, then given the light the finished world will throw on the performers: warm footlights for a music hall, an orange glow from a dragon that doesn't exist yet. Real light on real faces is what sells the shot.
- 06
Two ways to finish
Route one cuts the performer out, renders the world through the camera's real move, lets a model dress it and puts the performer back untouched. Route two gives a model the shot plus a 3D layout and it paints the world around them.
Three shows, one stage
Variety. A magician, a cabinet and a grey room become a period music hall. The crosses hold the track, so the camera can turn to face the audience and the whole hall turns with it. The trick and the performance happen on the day; the room and the crowd come later.
Comedy. A planning officer measures a dragon, or rather a grey wall. The phone's LiDAR measures the room in real metres, so the dragon is built to the right size in 3D, seen through the camera's real move, before it's dressed.
Vertical drama. Shot on a phone that films and tracks itself, with a grey sheet, a lamp and two chairs. Every floor of the lift is the same shot; when the doors open, a lamp lights the faces in the colour of the next world. Each test cost about £2, so you can try forty floors and keep the five that work.
The small print
Everyone on screen in Live Grey is AI-generated. Shadows are still the hardest part, and hair comes next. Models can forget a room if the camera looks away from it for too long. A recognisable face only goes in once that person has verified and agreed: a rule, not a setting.
It suits period drama, historical reconstruction, comedy and entertainment formats, vertical series, branded films and live events: anywhere the world is too costly, too far away or gone.
Drag to compare
The same take, before and after. Left of the line is the greybox plate; right of the line is the finished frame. Drag the handle, or focus it and use the arrow keys.
Workflow demo · all images AI-generated
Six steps, from the grey room to the edit
Workflow demo · all images AI-generated
- 01
Greybox
The performance is filmed for real in a grey studio, with a few flats and some tape on the floor. It's lit flat, with one even overhead light, and relit in post.

- 02
Frame one
The first frame becomes the design stage. The world is built on that one still, from period reference, and signed off before anything moves.

Before 
After - 03
Depth
A depth pass maps where everything sits in the room.

- 04
Transform
The clip goes through the model and comes out in the new world, with the actor's timing and the camera move intact.

Plate 
Final - 05
Sign-off
Every shot reaches the producer as a still first. Anything that doesn't work is redone before it's animated.

- 06
Delivery
Approved shots go straight into the edit with notes attached, ready to cut.

Sheets and approved keyframes
Characters and locations are drawn up as sheets before any shot is made, so every frame is built from the same references. The seven keyframes below were approved as stills before they were animated.
Workflow demo · all images AI-generated











Stand where it stood.
Watch it come back.
The Crystal Palace burned down in 1936. This teaser starts in the park at Sydenham today, rebuilds the glass palace as it was in 1899 from archive reference and fills it with visitors. Halfway through it stops to show how it was made, from the evidence to the true-scale model to the finished frame, then brings back the night it burned.
It shows how a history series can put a presenter in a place that no longer exists, and move between then and now in a single shot. The presenter here is an AI stand-in.
AI-generated images and 3D model renders · presenter is an AI stand-in
A period reconstruction teaser: the Crystal Palace, Sydenham, 1899 and 1936.
Rebuilt from archive reference · 2 min 6 secBuilt to scale before a frame is made
Nothing of the building survives to film, so the work starts with the record. The main dimensions come from period sources. The palace is then modelled at true scale in Blender, so every camera sees a building of the right size and shape.
- Length of the building
- 1,608 ft
- Height of the central transept
- 168 ft
- Height of the nave
- 104 ft
- Height of the water towers
- 284 ft
Blender model · true-scale blockout




Four cameras, from block model to finished frame
Each camera gives three guides: a block render, a depth map and a line drawing. The finished frame is made on top of them, so the architecture stays the same from shot to shot.
Guides rendered in 3D · finished frames AI-generated
















Five steps, from the record to the edit
- 01
The evidence
Plans, measurements and photographs of the building as it stood.
- 02
Built to scale
A plain 3D model at true scale, with cameras placed where a crew would stand.
- 03
Checked
The model is checked against period photographs. Where they disagree, the photograph wins.
- 04
The finished frame
Glass, ironwork, gardens and visitors in the dress of the time, signed off as a still.
- 05
Then and now
Approved frames are animated and cut against the park as it is today.
TV is changing channels.
Everyone's coming along.
The old way and the new way, in one gondola. Two character sheets, one location and a fight that returns to its opening frame. It's a 28-second loop built through the same workflow we bring to a series, where every character has to look like themselves in every shot.


From the script
to the sofa.
This is the production workflow behind the scroll film on our homepage. One script goes in at the top. What comes out is a cut, graded programme on a living room telly, with every shot traceable back to the page.
- 01
The brief
Every script is broken into shots, references and a continuity passport before anything is generated.
- 02
The board
Our platform tracks each frame from prompt to approval, so the edit always knows what it's cutting.
- 03
The world holds
Same rooms, same faces, same light, episode after episode.
- 04
On air
Cut, graded and on air.
The machine lays the cut.
The editor makes the film.
From the rushes to a first cut in an evening, in our own edit suite.
Every word in every clip is transcribed and logged. Claude reads it all against the brief and writes the edit down, with a reason for every choice. Then CC Edit, our edit suite, lays the cut from those instructions: sixty-three of them, in under a second.
Notes on a shared cut come back as a new version on the same link. Every shot is checked against the rushes it came from, and every AI tool is logged for the commissioner. The film shows the whole line on STUDIO 4, a short documentary we cut this way from fifty AI-generated rushes.
Edited in CC Edit · STUDIO 4's rushes are AI-generated · narrated by an AI voice
Cut by Instruction: the edit workflow, with STUDIO 4 as the proof.
The lay captured from CC Edit · 4 min 17 sec- 01
Logged
Every word in every rush is transcribed, so the footage can be searched like a script.
- 02
Written down
The paper edit is written against the brief before a frame is laid. The brief to the first cut took twenty-one minutes.
- 03
Laid
CC Edit lays the cut by instruction, and every shot is checked against its rushes.
- 04
Noted
An executive's notes came back as a new version nine minutes later. The editor still decides what the film is about.