The Shot Prompt Carries 9 Blocks

The template is 1 page long, and it never changes shape. 9 blocks, in this order: identity, subject and action, composition, lens, lighting, movement, mood, quality anchors, negative. Written in that order every time, the prompt stops being a wish and starts being a specification. Reorder 2 of the blocks and the model reads a different film. Put mood above lighting and the light gets guessed instead of named. Each block answers 1 question and no block answers 2. Who is in frame. What happens. How it is framed. Which glass. Where the light falls from. How the camera moves. What it feels like. What it must look like. What it must never look like. 9 answers, 9 lines, no negotiation between them. A prompt that drops 1 block is not a shorter prompt. It is 1 decision handed back, and the model will take it every time.

8 shots carry the whole film, and each 1 is asked for 1 thing. Shot 1 is an empty studio at dusk. Shot 2 is her arrival. Shot 3 is the teacher in the doorway. Shot 4 is the first touch. Shot 5 is the missed step. Shot 6 is the distance closing. Shot 7 is the knowing. Shot 8 is the held moment. 1 beat per shot is the rule the file is built on, and the reason is arithmetic rather than taste. A shot that asks for 2 actions returns neither. The model averages the 2, and the average of 2 beats is no beat at all. So a scene that reads as 5 beats on the page becomes 5 shots, and each shot is written to 1 of them. The cut is where the extra beats live. Nothing inside a frame has to move for the scene to build.

2 identity blocks sit at the top of the file, and both are copied word for word into every prompt where that character is in frame. Her block names the character, her age, her hair, her skin and her wardrobe in 1 line. His block does the same in 1 line. Same face throughout is written into the prompt rather than hoped for across 8 separate generations. Change 1 adjective in that line and the face drifts. The drift is small in 1 frame and unrecoverable across 5. So the block is the only part of the template that is never rewritten, never shortened and never paraphrased from memory. Everything around the identity block changes. The block does not move 1 word.

4 lenses run the film: 24 mm, 35 mm, 50 mm and 85 mm. They are assigned per beat-group rather than per shot. Shot 1 holds 24 mm for the room alone. Shots 2 and 3 hold 35 mm for the 2 arrivals. Shots 4, 5 and 6 hold 50 mm. Shots 7 and 8 hold 85 mm. A beat-group shares 1 lens, and that is what stops a cut from reading as a jump. 2 shots at the same focal length cut together cleanly even when the framing changes underneath them. Swap 1 shot to a different lens inside its own group and the seam shows immediately, before anything about the frame itself is judged. 1 lens per group, 4 focal decisions made once rather than 8 made on the day.

9 blocks in a fixed order, and the order is the method.

Every lighting block names where the light comes from and which way it falls. Amber window light from the side is a source and a direction. Warm light is neither. One can be lit again tomorrow and the other cannot. Wherever a face is lit, the key sits on the far side, so the near cheek falls away and the face keeps its structure. The quality anchors then close the block with 6 words: documentary film, 35 mm, natural skin texture, visible pores, unretouched, negative fill. Those 6 do the work that the word cinematic cannot, because cinematic is a mood and pores are a measurement. A face that glosses over is not a lighting error. It is a missing word.

The last block is the longest and the cheapest. The negative line names 19 traps in 1 line. 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. 19 items, 1 line, copied into every prompt in the file. The line costs nothing and saves 1 render every time it fires. Then the arc, which is 1 move from start to finish: wide to close as it rises. 24 mm at the top, 85 mm at the end, and no reversal anywhere between them. 8 shots, 4 lenses, 2 identity blocks, 9 blocks per prompt, 19 negatives. That is the whole specification, and it fits on 1 page. The physical builds ran on the same shape. Companies built across 12 countries. 210 energy systems across Africa and Asia. 200 wood-gasification machines across the UK and Europe. Not 1 of those rows was improvised on site. The specification came first, and the build was the commodity that followed it. The probability that 8 separately prompted shots assemble into 1 continuous lesson is not 0. It is just low enough that the film stops being a scene and becomes 8 clips in a folder. So write the 9 blocks once, in order, and let the template carry the joining.