How the Red Ferrari Night Held Together
Thirty-five frames of one night in 1979, and the part that actually makes them agree with each other: the cast and the rooms, defined once and wired into every shot that needs them.
Thirty-five frames, one night, a seven-year-old and his father. The boy has to be the same boy in all thirty-five. His father has to be the same man. The club has to be the same room from the door to the bar to the backstage hallway. That means the bar you glimpse over a shoulder in frame 19 has to be the bar you are standing at in frame 22.
That is the whole problem. Not making a good image. Making the thirty-fifth image agree with the first.
Start with the script
The scene came out of Chapter One of a novel I have been failing to finish since 2023. The chapter is fiction told in the narrator's voice: Mason is seven the night his father turns up at his grandparents' house in a brand new Ferrari, red and gleaming under the streetlights.
That is not shootable. So it gets adapted into a screenplay: scene headings, shot IDs, and reference tables for the cast, the settings, and the style.
### SCENE A-1: GRANDPARENTS' PORCH - NIGHT
**WIDE SHOT** establishing a modest Grant Park Victorian. Warm yellow
light spills from windows onto a manicured lawn.
| ID | Preset | Description |
| A1-1 | establishing | Victorian house at night, warm window light |
| A1-4 | close-up | Mason's face, red taillights in his pupils |
Thirty-five shots for the act. Each one gets a framing preset: establishing, tracking, two-shot, extreme-close-up. Framing is a decision you make once, in writing, rather than thirty-five times in a prompt box.
Then the graph builds itself
The screenplay opens in a panel beside the canvas. Build Graph reads the tables and lays out the scene: every character becomes a node, every setting becomes a node, and every row in the scene's shots table becomes its own wired lane that runs shot into action into output, with the camera and the era wired into each output's config.
It works a scene at a time, and what it hands you is not finished. It is scaffolding: exactly what you want before the interesting decisions start.
You could render straight off the scaffold. In practice, this is where the work moves into the graph: tightening a shot, swapping a preset, rewiring a reference until the tree gives you the frame the scene deserves. Build Graph is a get-everything-started-for-me button, not a one-click-and-done button. What it takes off your hands is dragging boxes around, the part that never deserved an afternoon in the first place.
The cast is the actual work
Here is where consistency is won or lost, and it is not won by prompting harder.
Each character gets defined once, concretely, with wardrobe and posture instead of adjectives. Then you generate one clean full-body portrait on a plain background, and that portrait becomes a reference image wired back into the character node.
From then on, every shot that pulls that character pulls the picture too. The model is not being asked to imagine a seven-year-old boy thirty-five separate times. It is being handed the same boy, thirty-five times, and told what he is doing.
That is the difference between a tool that makes images and a tool that makes a sequence. The face does not drift because nobody re-described it.
Settings get the same treatment. The club got a clean empty-room plate, with no people and no action, just the room with its stage and its bar and its string lights. Wire the plate into a shot and the club stays the club.
Pages, then a screening
Frames become pages, with panel layouts, per-panel crops, and gutters. The crops arrive as live sliders, so a panel that is cropping wrong is a drag rather than a regenerate.
Then the whole act goes into Cinezine, where it becomes something you can hand to someone: cut the frames that do not earn their place, reorder, caption, pick a theme, and export one HTML file that opens anywhere with no server and no account.
That file is Episode A. It is the same file whether you are reading it on a laptop or I am putting it on a screen in a room.
What it cost
Most of the wall-clock time went to the references, the portraits and the setting plates. Almost none went to the shots themselves, because by the time the references were wired the shots were mostly bookkeeping.
This is something I did not expect when I started building this. The work is not at the end. It is all at the front, in deciding who these people are and what the settings look like. After that, the frames come cheap, because every one of them draws on the cast and the rooms you already built.