AI Video
AI Film Work
I made every film and shot on this page myself, start to finish. Storyboards and character, set and prop references come from mico, my own production tool; prompts are written with Claude and mico; stills are made with gpt-image-2 and the video is generated in Seedance.
Finished films
Commercial · Lightwear, an in-house T-shirt brand
David and Goliath
1020 BC, the Valley of Elah. The soldiers are handing out TEAM GOLIATH buttons. One stone later, everyone is quietly taking them off.
mico · Claude · gpt-image-2 · Seedance
Trailer · In-house feature documentary
Jesus Calling: Investigated
The trailer for a feature documentary on where the book Jesus Calling came from. Period scenes carry the question: a typewriter under a desk lamp, a handwritten page, two women listening in a sitting room, a letter burning. Live action and AI are cut together; most of what you see is AI.
mico · Claude · gpt-image-2 · Seedance
01 / 04
Title sequences
An opener lives on a planned camera move, and generated video tends to drift off it. For two of these I blocked the scene as a plain 3D model in Unreal Engine 5 to lock the camera path, then generated the shot along it, so the move is designed rather than left to chance.
Show openerOpener for a client's interview show: title letters swing into place over a dark stage, then the cast walks on and holds their marks.mico · Claude · gpt-image-2 · Seedance 
The tapestryA slow camera move along an embroidered tapestry of The Emperor's New Clothes, its path locked with a 3D block-out first.mico · Claude · gpt-image-2 · Unreal Engine 5 · Seedance 
The mapThe camera travels across an old map as figures of golden light rise from it, on a path locked with a 3D block-out first.mico · Claude · gpt-image-2 · Unreal Engine 5 · Seedance
02 / 04
Film scenes
A film needs more than one good frame: the same faces across cuts, people who keep their marks, hands that handle real props, a camera that moves like someone is carrying it. These scenes test each of those, from a single close-up to a multi-shot sequence. Four come from the Jesus Calling documentary; two are in 3D-animation styles, one leaning realistic, one cel-shaded.
Morning kitchenA woman packs two lunches and talks as a boy runs past; hands, props and timing stay believable.mico · Claude · gpt-image-2 · Seedance 
The phone callFrom Jesus Calling. A woman listens on the phone; small shifts in the eyes and mouth carry the scene while she barely moves.mico · Claude · gpt-image-2 · Seedance 
Quiet timeFrom Jesus Calling. A scene opener with many people and many actions: a woman prays by the window, a room kneels and writes.mico · Claude · gpt-image-2 · Seedance 
Above the lakeFrom Jesus Calling. A man in a period coat walks out above a lake, then looks up; framed like a location shoot.mico · Claude · gpt-image-2 · Seedance 
The visitorFrom Jesus Calling. A visitor climbs to a doorway; the two women, the house and everyone's position hold from close-up to wide to interior.mico · Claude · gpt-image-2 · Seedance 
Snow ridgeA woman walks a snowy ridge at dusk while the camera follows handheld, like an operator on foot.mico · Claude · gpt-image-2 · Seedance 
The cellRealistic 3D-animation style: an old prisoner tosses and catches a ball in a shaft of window light; the throw, arc and catch stay physical.mico · Claude · gpt-image-2 · Seedance 
The trenchCel-shaded 3D style: a drone drops a handset into a war trench, and one soldier's face holds from wide to close-up.mico · Claude · gpt-image-2 · Seedance
03 / 04
Interview show
An interview reads false the moment a face and its voice disagree. Here the voices were recorded in a studio first, then the picture was generated to them: lips in sync, gestures that fit the words, close enough to pass for a live shoot. Same studio as the show opener above.
The answerThe guest answers an off-screen host in a studio-recorded voice; the picture is generated to it, lips and gestures in sync.mico · Claude · gpt-image-2 · Seedance 
The breakMid-interview the tears build and she folds forward; the emotion grows across the take instead of switching on.mico · Claude · gpt-image-2 · Seedance
04 / 04
Emotion control
The hardest note to hit is a change of feeling inside one short shot. In each of these the emotion turns within five seconds and the face stays the same person throughout.
Through tearsShe cries, then a small smile breaks through as she wipes her eyes.mico · Claude · gpt-image-2 · Seedance 
The turnShe turns to camera, lowers her eyes, then looks back up with her mind made up.mico · Claude · gpt-image-2 · Seedance