Most product films on this page are written as timed shot lists. This one is not, and the difference is worth studying, because it is the better choice more often than people assume.
The four shots are marked by transition language rather than by timecode: "tight close-up on", "quick cut to", "final shot". That gives the model the sequence and the count while letting it distribute the eight seconds itself. In a film with this much texture — dials, mist, wet asphalt, print grain — the model needs that freedom, because the amount of time a shot needs to read is a function of how much detail is in it, and the prompt cannot know that in advance. Templates further down this page do use hard timecodes, and they are the right call when the picture has to hit an external beat. Here there is no external beat.
The sound is the actual subject. Read the prompt again and count how much of it is audible: a shutter click that echoes, a thumb wheel that is described as tactile, rain that has already fallen and is still dripping off everything. In a silent model none of that would survive — you would get a nice-looking camera. H3 generates audio in the same pass as picture, which means the words "a deep shutter click echoes" produce both the sound and the frame where the sound happens, and the cut lands on it.
That is the mechanism worth taking away from this template: on a model that generates sound, naming an audio event is also a way of scheduling a picture event. The prompt never says when to cut to the pedestrian. It says the shutter clicks, and then it says cut. The click is the edit point.
The monochrome instruction is load-bearing rather than stylistic. Black and white removes colour contrast from the model's toolkit, and what is left is shape, reflection and grain — which happen to be exactly what a wet night street is made of. It is a case of a constraint that improves the output by narrowing what the model is allowed to solve with.
The closing shot is the only part of the film that is an argument rather than a description. A stack of monochrome prints beside the camera says the camera works, which is the one thing a product film has to establish and the one thing a beauty shot of the product never does.
The two negative instructions at the end — "no text, no logos" — plus "fully unbranded" earlier are not redundant. The first pass suppresses the badge on the body; the second suppresses the model's urge to put a brand name on the prints.
Video