Sam Clarke · AI Cinematography
Untitled #1
A generative film sequence. Midjourney, Kling, DaVinci Resolve.
24 seconds · three shots · sound
Frames
01 · ext. wheatfield · duskFour figures head down through the wheatfield, the coast below. Sun just gone.
02 · mcu · duskHe has stopped. The others keep walking. Last of the light on his face.
03 · ext. country lane · duskBehind a worn white van on a narrow lane, deep into twilight.
Approach
Every setup is written like a brief to a camera department. Subject and identity, composition, light, mood, grade. The model gets exact instructions, not inspiration.
Continuity is the hard part, so that is where the time goes. Casting, wardrobe and light are locked in language and restated word for word between shots, which is why the coverage cuts together.
The rules stay strict across a project. Natural light only. Reserved performances. Social realism, not futurism.
The closeup was written to match the wide: same jacket, same sky, same last light. Different shot, same film.
Pipeline
- Look development
- Midjourney. Moodboard-referenced, prompts written like camera department documents: stock, lens, light, blocking.
- Cast & continuity
- Character and style references held across setups. Costume language kept verbatim between prompts.
- Motion
- Kling, image to video. The still carries the look; the prompt carries only the movement.
- Finish
- Cut, graded and grained in DaVinci Resolve. One pass across every shot so the sequence reads as one film.
- Grounding
- I shoot 16mm on vintage Arriflex cameras I bring back to working order. Knowing what photochemical texture actually looks like is most of the battle.