08:40 on a Thursday. The overnight batch has finished and there is a folder of frames waiting. I do not open it yet. Brief first, frames second, always in that order, because meeting output before you have reloaded the intent is exactly how a set drifts.
Nobody here carries the title prompt engineer. The title is creative director and it means what it always meant. Somebody is responsible for whether the thing is any good.
What follows is the shape of a real day, not a manifesto about one.
09:00 — Read the brief before the batch
The first hour is calibration, not creation. For Doré it means rereading why pleated metal goes cold on white, and why the still life sits on things that weather instead: a backlit banana leaf, a river stone, an open lily, raked sand, driftwood. Then the folder opens.
Frames look better than they are when you meet them cold. The average output of a good engine is competent, and competence is precisely the thing that slips past you at nine in the morning. Rereading the direction gives you something specific to fail against.
Three piles: keep, fix, kill. Most of it is kill. That is not the engine failing — range is what the engine is for, and range means you asked for material you were always going to discard.
11:00 — Briefing, which is not prompting
Mid-morning is briefing. A prompt describes an image. A brief describes a decision, and the difference shows up about two hundred frames later.
RIIFFS runs two lines under one roof. NUSUK carries Arabian oud and heritage; RIIEFF takes the Western-styled designer side, with Samah and Privé inside it. The bottles are shot in studio and the world behind them is generated, then composited in. Generative intelligence will build that world in minutes. It has no idea which of the two worlds it is supposed to be building.
So I write constraints rather than adjectives. Shadow does most of the work. Nothing behind the bottle. A constraint survives the trip into a model, back out into a compositor, and into a junior's hands at eight in the evening. An adjective does not survive the first handover.
Half of briefing is deciding what would make a frame wrong. That sounds like a negative exercise and it is the most productive hour of the day.
14:00 — The review, out loud
After lunch the set goes up on one wall, at one size, in one pass. No flattering thumbnails.
Two rules govern the review. Every frame is judged beside its neighbours, because a house is a set and not a run of strong singles. And every kill gets its reason said out loud. The gold is too orange, it reads sweet. The stone is doing nothing. Speaking it teaches the room; a silent delete teaches nobody.
L'Orientale Fragrances is the clearest test we have of that discipline. The same bottle has to hold on a dark cinematic background, gold and amber on matte glass, and then again on plain white for the product grid. If the direction only survives one of the two, it was never direction. It was a look.
Reviews run long and they are the cheapest hour in the building. An unspoken objection at two o'clock becomes a rebuild of the world at six.
17:30 — Finish, where the day actually goes
The last block is finishing. Colour, light, the small corrections nobody itemises. It takes longer than everything before it combined, and no client has ever asked for it by name.
On COLOSSEUM, the film for Christian Provenzano, the composition runs against classical architecture and frames the house on the perfumer rather than the product. The engine gave us the range to try that at all. It could not tell us the film belonged to a man instead of a bottle. Somebody in a room decided that, and then spent the grade defending it.
18:20. Tomorrow's batch is queued and it will run overnight whether or not I have a point of view about it. Infinite output against finite judgement — that asymmetry is the entire modern job of a creative director in an AI studio, and the day is only its timetable.