The Frame Is Written Before The Motion

This sheet is 3,105 bytes, its sha256 hash begins 0345ecaac6f38020, and it holds 12 shots numbered x01 through x12 with no gap in the run. Each shot carries exactly 2 fields. FRAME comes first and MOTION second, and the header states that order before any shot appears. 12 shots, 2 fields each, 24 fields, 60 lines. Of those 60 lines, 16 are blank, 1 is a rule, and 12 are shot titles. The remaining 24 lines are the fields themselves. The ids are contiguous and the field count is constant, so the sheet reads as a table by eye or by a script with no special case anywhere in it. The sheet also names both characters once, in 22 words, at the top, so no row repeats them. A sheet built this way is parsed, not interpreted.

The 12 FRAME lines carry 212 words: an average of 17.67 each, a shortest of 14 and a longest of 21. The FRAME is the still, generated as a 16:9 image or taken from a reference still, and the header says which. Every one of the 12 frames is a single sentence, so no frame makes 2 arguments. The vocabulary is decisions rather than praise. The word artistic appears in all 12 frames. Warm candlelight appears in 8 of them, silhouette in 5, backlit in 4, tasteful in 3, and 2 frames state outright that no face is in the picture at all. The same word artistic appears 0 times in the 12 MOTION lines. That is the split the sheet is built on: the words about how it looks stay in the frame, and the motion column gets verbs.

The 12 MOTION lines carry 66 words: an average of 5.5 each, a shortest of 3 and a longest of 7. Against 212 words of frame that is a ratio of 3.21 to 1, and the relationship holds on every row of the file. Not 1 of the 12 shots has a motion longer than its frame. The shortest frame, 14 words, is still twice the length of the longest motion, 7 words. The most common verb in the motion column is slowly, used 5 times, and closer appears 3 times. The pace is stated rather than left to the model. Motion from nothing has to describe a world. Motion from a fixed frame only has to describe the change, which is why 5 words can be enough.

The shared style suffix is 33 words, 30 of them unique, 223 characters, and it is appended to every prompt rather than retyped into all 12. It carries documentary style, a warm summer night, humidity, amber candlelight and soft moonlight, deep shadows, film grain and a shallow depth of field. It closes with 4 prohibitions, every one of them about what must not appear in the frame. Retyped into all 12 prompts that suffix would be 396 words. Written once it is 33, which is 6.86 per cent of the 481 words in the sheet. The look is a constant and the shot is the variable, and the file says so on the line above the first shot. A look typed 12 times has 12 chances to drift, and the drift lands in the last shot of the day.

3,105 bytes. 60 lines. 12 shots.

Image-to-video takes a still and a motion line. Whatever the motion line says, the picture underneath it was already decided by the frame. That is why the order is load-bearing, and why FRAME is the first line of all 12 rows. A motion written before its frame describes a scene that does not exist yet, and the model resolves that gap by inventing the frame quietly. What comes back looks like a render of a prompt rather than a shot of a plan. The 2 fields are not halves of the same task. They are 1 decision and 1 consequence. The frame spends the money and the motion spends the time. One is 212 words and the other is 66, and that gap is the whole difference in what each field is allowed to choose.

The probability that 12 shots, generated weeks apart by different hands, would land on the same light, the same shadow depth and the same grain without the look being written down once and reused is not a number worth betting a campaign on. The suffix is the cheap insurance: 33 words that stop 12 prompts becoming 12 looks. Of the 481 words in the sheet, 278 are the FRAME and MOTION rows themselves, 57.8 per cent of the file spent on the 2 fields that reach a camera or a model. On the 210 energy systems deployed across Africa and Asia the rule was the same: each unit ships with the specification of the unit that was tested, because a rebuild from record and a rebuild from memory are 2 different objects. A row that carries its frame and its move is a shot that can be repeated. A row that carries only a title is a shot that was witnessed. The frame is written before the motion. Keep more of yourself, for longer.