Status: APPROVED by Mabel, 7 August 2026. This is the film's photographic canon; the scene sheets inherit from it rather than restating it, and a prompt that contradicts it is a defect. It exists because the film had acquired a photographic style nobody signed off: every prompt sent today opened with *"Unretouched documentary photography, shot on 35mm film with visible grain" and closed with "deep focus"*, and neither phrase appears in any sheet, in `PLAN.md`, or in any decision. Meanwhile the BOILERPLATE that is approved says "shallow depth of field." Two opposite instructions were nominally in force and the undocumented one was winning, because it was the one being pasted into prompts.
That is the gap this closes. The recommendations below were approved as written on 7 Aug 2026; the four open questions at the foot remain open.
Three things about The Offer should decide its look before any taste does.
It is a film about work with materials. Ink, damp paper, oak, iron, leather, lampblack. The presses are period machines whose mechanism the audience has to read — the bar coming over, the platen biting, the sheet peeling off wet. A look that dissolves the machinery into bokeh is photographing the wrong film.
It is a film about a refusal, staged in one room. The drama is faces at a table deciding things, then a man saying no. That wants proximity and patience, not coverage and movement.
Its climax is a number on a page. The 400-gulden figure is withheld until the final shot. The whole film is built to make one insert land — which means the grammar of inserts has to be established early and kept scarce, or the last one is just another close-up.
The temptation is 2.39:1 for period grandeur. Refuse it: `seedance_2_5` tops out at 720p, so cropping to scope leaves 1280×536 — throwing away a third of the only pixels we get, on a film with no landscapes to justify it. 16:9 also suits a story of interiors and faces. If a wider frame is wanted later, it is a grade and matte decision on a finished cut, not something to pay for in resolution now.
Keep the 35mm framing, but qualify it, because "visible grain" is what has been sent and it is too strong an ask. Recommended wording: **fine 35mm grain, present in the shadows and the sky, never crawling on skin.** Two reasons. The film is about texture, and grain is the photographic argument that these are real surfaces rather than renders. But at 720p, heavy grain competes with actual detail and wrecks compression — the noise eats the bitrate that should be describing wood grain and wet ink.
This is also our main defence against 2.5's gloss. Reviewers consistently describe the model as polished and smooth; every take today needed "unretouched documentary" doing real work at the top of the prompt. Keep that phrase. It has earned its place empirically, which is more than the rest of the style could claim this morning.
This is the contradiction to settle, and the answer is that both were right about different shots:
The payoff is structural. If shallow focus means *"look, this object is the evidence,"* then it is a grammar, and the final insert on the number inherits every previous use of it. Shallow focus everywhere would spend that for nothing.
Naming a focal length in the prompt is also a practical lever: it constrains the model's default, which drifts wide and distorted at the edges.
Chest height, level horizon as standard (already the sets' convention). Movement should be limited to what a person could do: a slow dolly, a pan, a handheld follow. No crane, no drone, no orbit, no push-in that accelerates — these are the model's favourite flourishes and they are anachronistic in effect if not in fact, because they announce a camera the 1539 world cannot motivate.
One deliberate exception: the runner's exteriors, which track low and fast because the sequence is kinetic by design. There were two until 7 Sep 2026 — the second was M-13's slow pan across a shop going quiet, where the move was the beat — but M-13 was cut that day and the exception goes with the shot rather than looking for a new one to attach to.
Note that the rule above does not need bending for the over-the-shoulder push now under consideration for M-12. It is a slow dolly at roughly 0.3 m/s that holds, travels and settles without accelerating, which is squarely "what a person could do"; the clause it would fall foul of is *push-in that accelerates*, and it does not.
Today's most expensive lesson. Describing action at millisecond granularity — leather creasing, a sole skidding, grit hanging in the air — reads to the model as a request for high-speed photography, and M-01 came back at roughly quarter speed. State the cadence as a fact in any prompt with fast motion ("four footfalls every second, each foot smeared with motion blur"), and keep micro-detail in the costume description where it is not asking the camera to slow down to see it. Standard 180° shutter; motion blur present and normal.
Inherit what the approved sets already are: cream-ochre lime plaster, warm brown oak, terracotta, iron grey. Then one rule that carries the whole film: **the warmth is the sun's, and nothing else is warm.** Morning exteriors and the shop are golden where light falls and blue-grey in shade; the evening council is one candle in a cold room; the study is neutral daylight.
Avoid the teal-and-orange separation the models reach for by default. Blacks lifted slightly, as film shadows are, rather than crushed to digital black.
The film's device is that the number is hidden until the last shot. Recommend the final insert be the only shot that breaks the established grammar — whether that is the sole static locked-off frame in a film of moved cameras, the longest hold, or the only true macro. Reserve one photographic gesture, use it once. The audience will not name it, but they will feel the film stop.
1. 35mm grain — keep, soften, or drop? My recommendation is keep and soften. 2. Deep focus as the default, with shallow reserved for the four inserts? 3. Do we commit to the three-lens set, or leave focal length unspecified and accept the model's default? 4. What is the one formal break at the end — static hold, longest take, or macro? This is the only recommendation here that is a directing decision rather than a photographic one, and it should probably be made against the locked script rather than in the abstract.