Personal Project · Analog Photography × AI × VFX

Still, Moving.

Analog stills, brought to life.

Two stills shot on black-and-white film, set in motion with AI, without losing a single grain of their analogue character.

Project
Analog Photography × AI imagery × VFX
Photography & AI
Robert Brembeck
Compositing
GOLD Creative Studio
Camera & Film
Leica M6
Kodak Tri-X

The project

The base images were shot on film: a Leica M6, almost 40 years old, with a Leica Summicron lens from the 1970s, loaded with Kodak Tri-X, on Munich's Theresienwiese.

Into these real photographs we inserted AI-generated performers – a dog racing towards the lens, a skater taking her bow – seamlessly, as if they had always been part of the picture.

The idea

What fascinates us is how the unmistakable look and feel of black-and-white film, the grain, the tonality, the softness, carries over into the digital moving image.

The films.

01 · The dog. Circus fairground, Theresienwiese.
02 · The skater. Skatepark in the fog.

Insights

How we did it.

Generative video can animate any photograph. What it usually can't do is leave the photograph alone. Every frame is re-rendered: the grain changes, the tones drift, the architecture softens. For an analog image, that's exactly what makes it valuable.

The performances were generated with AI video from the original still as the starting frame. Then the real work began: getting the photograph back.

01

Clean plate from the negative

The camera never moves, so the original scan serves as a perfect clean plate. Every pixel of background in the final film is the photograph itself, not an AI re-interpretation.

02

Moving mattes

A frame-accurate alpha matte of the performer, extracted from the generated footage with a dedicated alpha-generation model – hair, cap, board and flying limbs included.

03

Shadows and reflections

A person-only matte leaves the subject floating. A difference matte, limited to a search zone below the subject, brings back contact shadows and reflections on the damp ground.

04

Atmosphere

In the skatepark, a depth-aware fog veil sits over the skater: denser while she is far away, thinning as she approaches the camera.

05

Tone match

The generated performers came out a touch brighter than the film. Frame by frame, a tone curve is measured from the generated background against the scan and applied only to the performer and the hand-over frames – the photograph itself is never graded.

06

Grain, measured and rebuilt

A still photograph has frozen grain; a moving subject has living grain. We measured the grain of the original scan, removed the frozen grain and laid fresh, animated grain with the same character over the entire frame.

07

Hand-over

When the dog jumps at the lens and the camera shakes, the plate no longer fits. The composite hands over to the generated footage over a few frames, unnoticed inside the motion.

Breakdown skater: original analog scan, generated performance, matte and shadow, final composite
Skater, generated frame
Generated performance
Skater, matte with shadow
Matte + shadow
Skater, final composite
Final composite
Breakdown dog: original analog scan, generated performance, matte and shadow, final composite
Dog, generated frame
Generated performance
Dog, matte with shadow
Matte + shadow
Dog, final composite
Final composite

The compositing steps run as our own nodes inside our local ComfyUI pipeline, so the method is repeatable for any locked-off shot: archive photography, campaign stills, key visuals that need to move without losing their original look.

Photography & AI Imaging: Robert Brembeck · Compositing: GOLD Creative Studio
See all projects