ai video production · lindenau · 2026
One scene, five worlds
An 18-second action sequence, shot once as a Western. Then re-dressed four times: new land, new light, new characters, new mounts. Every cut lands on the same frame in every world. Five shots run in all five worlds; the sixth, the escape, exists only in the wasteland.
the claim
The story is fixed. Everything you can see is a decision.
An anarchist sets a charge, a plainclothes officer fights him on the roof, five riders open fire, the train blows up short of the station. That scene exists once, as a Western. The four versions below keep its edit, its camera and its choreography to the frame — and swap the land, the century, the people and what they ride. All of it grows from one grey 3D previs; how, further down.An anarchist, a plainclothes officer, five riders, a train blown up short of the station — shot once, as a Western. The four versions keep its edit, camera and choreography to the frame and swap the land, the century, the people and what they ride. All of it grows from one grey 3D previs.
- source
- 18.0 s · 6 shots · 431 frames · escape in the wasteland only
- worlds
- 4 + original
- cast references
- 20 · white studio
- cut drift after conform
- 0 frames
- base
- one grey 3D previs
- image · motion
- gpt image 2 + nano banana 2.1 · seedance 2.5
- sound
- sfx only
shot board
Five shots. Five answers to the same frame.
Pick a world and the board re-dresses in place: the camera, the plunger box, the riders’ positions and the explosion stay where the director put them. “All five” lays the versions side by side, shot by shot.
frames taken from the conformed clips at the same timecode in every world
casting
The cast is written before the scene is shot
Every character and every special vehicle gets a casting clip first: full length, white studio, one slow blink. Those stills become references inside the video prompt, so a hooded shinobi stays the same hooded shinobi from the roof to the blast. The bodies in the source film are replaced; their movements are not. One screen per world: the world on the left, its cast and its machines on the right.Every character and vehicle gets a casting clip first: full length, white studio. Those stills become references in the video prompt, so a hooded shinobi stays the same shinobi from the roof to the blast. The bodies are replaced; their movements are not.
who decides what
What stays locked, what gets re-written
Control is the product. A client can approve the story once and ask for a different century without a reshoot.
Locked, frame for frame
- editsix shots, five cuts, 431 frames; the escape only in the wasteland
- camerathe low roof camera, the zoom onto the riders, the locked wide of the blast, and in the wasteland the pan after the escape
- choreographythe shove at 2.0 s, the plunger at 8.0 s, the flash at 10.3 s, and in the wasteland the mount at 14.6 s
- trainpath, speed, derail, where it stops
Re-written per world
- landfrontier plain · frosted village · green meadows · dead fields · blossom and paddies
- lighthard sun · dusk and lamps · overcast · broken sky · golden hour
- peopleanarchist and inspector · agent and enforcer · scavenger and warlord · shinobi and samurai
- mountshorses · horses · war bikes and buggies · horses with lacquered saddles
- weaponscarbines · revolvers · hand-built guns · bows
the base
Where it all comes from
One grey 3D previs, a set of master frames and a first photoreal film. Everything above grows from it.
the base · how it was built
Lindenau
The base of everything above: a steam train reaches a village station at blue hour. Camera, cuts and timing come from a grey 3D animatic. Every pixel of the picture is new.
what the previs decides
The animatic decides time and space. The picture is built on top of it.
Master frames set the look shot by shot, made with GPT Image 2 and Nano Banana 2.1, each model where it fits. Seedance 2.5 moves them with the animatic as its only motion reference. The previs locks edit, camera, blocking and the train’s action; the picture adds light, material, lens, atmosphere and sound (steam, wheels, wind, bells; no music, no voices). The result cuts exactly where the director cut, because the cut was never touched.Master frames set the look shot by shot (GPT Image 2, Nano Banana 2.1). Seedance 2.5 moves them with the animatic as its only motion reference. The previs locks edit, camera, blocking and the train; the picture adds light, material, lens, atmosphere and sound. It cuts exactly where the director cut.
- format
- 1920×1080 · 24 fps
- runtime
- 27.4 s · 7 shots
- master frames
- 8 on the board · 9 in the film
master frames
Eight frames, art-directed one by one
Before any motion, every shot got one still: its first previs frame, photographed again. Each frame used two references, the previs for composition and the style frame for light, lens and materials. Left: the exact previs moment. Right: the master frame that drove the shot.Before any motion, every shot got one still: its previs frame, photographed again with two references — the previs for composition, the style frame for light, lens and materials.
01 · village, aerial
← → frame · space a / b · f full screen
board
Before and after, shot by shot
Switch the whole board at once. Every frame stays in its place, so the change reads shot by shot. Pick a world and the same cuts re-dress; shots 08 and 09 exist in the final film only. Open any frame to see the pair full screen.Switch the whole board at once — every frame stays in place. Pick a world and the same cuts re-dress.
shots 02–07 are in the film · 08–09 extend the sequence · one master frame covers 07 and 09 · shot 01 is in the film above, not on the board
workflow
Four passes. One film, then four worlds through the same four.
Time and space first, picture second, motion last. Every pass says what it did for the film and what it does again for each world. We say plainly where AI works and where the studio does.
- 01 · studio
Previs
A grey 3D animatic. Camera paths, focal lengths, blocking and the train’s speed are locked here, cut by cut. The film and all four worlds run on this one animatic.
9 shots · 40 s · untextured - 02 · ai + studio
Look
- the film
- One style frame sets the world; each shot’s first previs frame is then photographed again with it as reference.
- every world
- Eight world frames from the same previs stills, each with a light report (kelvin, ratio, lens, grain), plus a casting still per character and vehicle.
- 03 · ai
Motion
- the film
- Seedance 2.5 in video-edit mode: the animatic drives camera and objects, the master frames define the look.
- every world
- The original scene as motion reference, three world frames as the law for light, the cast stills bound to each role by name.
- 04 · studio
Conform and finish
- the film
- Three shots regenerated, every cut trimmed four frames each side. Effects only, no music, no voices.
- every world
- Cut on the same five cuts, timed to the source shot lengths; frame sheets against the beat map, then a wording fix, a reference fix or a re-roll.
kurz auf deutsch
KI-Videoproduktion für Agenturen, Produktionsfirmen und B2B-Marken
Lindenau ist eine 18-Sekunden-Actionsequenz, einmal als Western gedreht und danach aus einer grauen 3D-Previs viermal neu eingekleidet: neues Land, neues Licht, neue Besetzung – Schnitt, Kamera und Choreografie bleiben in allen fünf Welten bildgenau. So arbeitet unsere AI Production Unit aus München: Previs in Blender, Cinema 4D oder Unreal, Master-Frames werden freigegeben, bevor etwas generiert wird, jede Einstellung wird Bild für Bild geprüft – für Werbespot, Produktfilm, Imagefilm oder Messefilm. Mehr zur Videoproduktion in München und Nürnberg.
- 01 · frage
Was ist ein KI-Hero-Shot und wann lohnt er sich?
Eine einzelne Einstellung, die klassisch nur mit CG, VFX oder einem Nachdreh zu bekommen wäre: ein anderer Ort, eine andere Epoche, ein Produkt in Bewegung. Gegenüber dem klassischen CG/VFX-Weg bis zu 5× schneller und um ein Vielfaches günstiger – bei Hero-Shots und CG-Sequenzen.
- 02 · frage
Was kostet eine KI-Sequenz?
Festpreis pro Shot oder Sequenz und Termine in Arbeitstagen – beides schriftlich, bevor wir anfangen. Ein Pilot-Shot zeigt vorab, ob der Look trägt.
- 03 · frage
Welche Tools kommen zum Einsatz?
Previs und 3D in Blender, Cinema 4D und Unreal Engine; Bilder mit GPT Image 2 und Nano Banana 2.1; Bewegung mit Seedance 2.5. Regie, Previs, Frame-Auswahl, Casting und Schnitt bleiben im Studio.
- 04 · frage
Für wen ist die AI Production Unit gedacht?
Für Agenturen und Produktionsfirmen, die Pitch-Frames, Previs, Hero-Shots oder ganze Sequenzen zuliefern lassen – auf Wunsch unter eigenem Namen – und für B2B-Marken aus Deep Tech, Industrie und SaaS. Von München aus in der DACH-Region und europaweit.
your project
Have a previs, a storyboard or a CAD model? We take it to picture — and keep every cut you approved.
Direction, previs, frame selection, casting and edit by Viatkin Productions. Previs and 3D in Blender, Cinema 4D and Unreal Engine. Images: GPT Image 2 and Nano Banana 2.1. Motion: Seedance 2.5. Sound: effects only.