The Axis Is Set Before The First Shot

The axis is 1 decision and it sits above every frame in the list. It is the 180 line, and it is the shortest rule in the document. The line runs along the dance floor with the 2 people facing each other on it, and all 3 camera positions stay on the mirror-wall side of that line. Cross it mid-scene and the 2 flip sides on screen, which an audience reads as 2 people who have swapped places; no cut repairs it, and no grade hides it. When the 2 turn together the axis rotates with them, so the rule is not a wall. It is a relationship that moves with the bodies. The probability that screen direction holds across the whole scene is settled here, in 1 sentence, rather than in the edit. The reason is mechanical. A viewer builds a map of the room inside the first 2 shots, and the axis is what keeps all 6 later shots on that same map. Break it once and the map is redrawn. Stay on the mirror-wall side and nobody has to think about where anyone is standing, which leaves them free to watch her face.

The lens is chosen 3 times for 8 shots and never mixed on the same subject. Beats 1 to 3 sit at 24mm and 35mm, because those 3 shots are about the room, the space and the distance between 2 people who have not touched yet. Beats 4 to 6 sit at 50mm, the natural lens, because those 3 beats are the dance and a lens should not be making an argument while 2 bodies are. Beats 7 and 8 sit at 85mm, which isolates and pulls the frame into her skin for the knowing and the held moment. The glass tightens once per group, 24 to 50 to 85, and 3 groups is enough resolution for 8 shots. Mixing a 35mm into the 85mm group costs the scene its progression; the cuts stop reading as approach and start reading as noise. The 3 decisions are made once, written into the table, and every later render inherits them. The recap is 1 line: 24mm and 35mm for 3 beats, 50mm for 3 beats, 85mm for 2 beats.

Identity is asserted in the list rather than hoped for in the render. 2 blocks are locked verbatim and prepended to every character shot: 1 for her and 1 for him, each naming the face, the hair, the age and the clothing in fixed words that never get rephrased. The first runs 27 words and the second 23. 7 of the 8 shots carry a character, so the same 2 paragraphs are pasted into 7 prompts without a word changed. That is the mechanism that holds 1 face across a scene on a model with no memory between frames. Rewrite the block, even to improve the prose, and the model redraws the person. The blocks also do the continuity work that a costume department would do on a live shoot: 1 charcoal practice dress, 1 black shirt with the sleeves rolled, and 1 face each, held for the whole 8 shots. 2 paragraphs, 7 prompts, and the face survives.

The table has 8 rows and every row carries 6 fields: the beat, the shot, the size and angle, the lens, the movement and the purpose. Row 1 is the empty room at dusk, locked off with a slow push-in at 24mm. Rows 2 and 3 bring the 2 people in at 35mm, static and then a slow dolly as she enters, over the shoulder for him. Rows 4 to 6 stay on the 180 line at 50mm, a lateral track with the dance and then 2 slow dolly-ins as the steps slow and the lesson stops being about steps. Rows 7 and 8 stay on that same line at 85mm, static, with the camera holding its side while the 2 of them stop moving. Rows 7 and 8 hold still at 85mm for the thumb that traces once along her spine and the moment neither of them lets go. A row that leaves the lens column open is not a row. It is a question, and 8 open questions is a scene nobody can render.

The axis is 1 decision and it sits above every frame in the list.

The list carries 2 global strings and they ride on every render, including the empty room. 19 negative terms name the failure mode in the model's own vocabulary: glossy, airbrushed, plastic skin, CGI, 3D render, digital painting, over-smooth, beauty filter, uncanny, doll-like, wax figure, cartoon, text, watermark, jittery, fast motion, extra limbs, deformed hands, waxy skin. 7 anchors name what the frame should be instead: documentary film, 35mm, natural skin texture, visible pores, unretouched, natural light, negative fill on faces. The 2 strings are pasted verbatim rather than re-typed, because a negative prompt rewritten per shot drifts, and the drift is what puts a wax face in shot 7 that was not in shot 4. 19 terms and 7 anchors, 1 pair of strings, 8 frames.

A shot list is the contract the render is held to, and it is written before a single frame exists. The source note here is 726 words, and the discipline is the same one that sets an axis at 24mm and holds it to 85mm. Hold 1 axis, choose 3 lenses, lock 2 identity blocks, fill 6 fields in 8 rows, paste 19 negatives and 7 anchors, and the scene has somewhere to stand when the renders come back wrong. Leave the lens open and the render decides for you. EX Venture runs the same shape across its studio: 35 staff and 200 plus interns a year, work built across 12 countries, 393 pages in the book and 450 plus generated articles and films behind the pipeline, and every render order passing a named check before it ships. The reason is identical in both places. A decision written down can be audited, argued with and improved. A decision that lives in someone's head lasts until the day they are tired. Set the axis, hold the glass, paste the blocks, name the negatives. Then shoot.