The Right Feeling, the Wrong Filmmaking: Should You Keep the Reference?
The Right Feeling, the Wrong Filmmaking: Should You Keep the Reference?
The image is on the board because it moved someone. Two people in a room that has been arranged rather than lived in, standing close, one of them handing over something small. It is a still from a short clip, and the still is the part that got pinned up. Whoever brought the clip in felt the thing the commercial is supposed to feel. It has survived three rounds of cuts and one budget conversation.
It is also, right now, proposing a different commercial.
The temptation is to keep it and write a caption that fences off the damage. That can work, but less often than people hope, and the reason is not mystical. A caption is one sentence competing against a few hundred decisions. When you keep a partial-fit reference, you are not choosing a mood. You are choosing which set of specific instructions most people in the room will receive.
So the answer up front: keep the reference only when its useful cue survives the misunderstanding its other signals will introduce. Not when a disclaimer has been added to it. Not when the argument for keeping it has gotten longer than the description of what it contributes. The useful cue has to still be doing work after you have subtracted everything the reference is saying that you don't want.
Everything below is a constructed example. I have invented the reference, the commercial and the annotations to make a method visible. No reader was tested, no collaborator was interviewed, and nothing here reports an observed result.
Name the feeling as a relationship, not an atmosphere
"Intimacy" is not a cue. It is a label for a result, and a label cannot be art-directed.
In the invented example, the clip shows two people close together in a glossy, highly controlled room. The desired quality is a hesitant exchange of a small object — a cup, a key, a folded card, it doesn't matter. What matters is that one person extends it and the other's hand opens without closing right away. There is a delay. The object is held for a beat longer than a handover needs, and then released.
That half-second is the reference's entire contribution to this job. Everything else in the frame — the room, the stillness, the finish, the styling — is either neutral or hostile.
If you cannot point at the moment in the clip that produces the feeling, you do not have a reference. You have a preference. And a preference cannot be given to anyone else to build from, because there is nothing in it to build.
List the film the image is also proposing
Now do the unpleasant part. Describe what the reference says when you stop reading it for mood and start reading it for instructions.
In the constructed example, the visible signals run something like this:
- Camera behaviour. Stillness, or a slow drift. Long holds. Shallow focus that isolates the two figures from the room behind them. Almost no cutting inside the moment.
- Casting and presence. Two people, composed, matched to each other. No one else in frame. Nobody is mid-task.
- Environment. A single sealed interior. Clean surfaces. No evidence of other people's work — no boxes, no cable, no one's coat on the wrong chair.
- Finish. Gloss and controlled falloff. Highlights that were graded to sit where they sit.
The invented commercial needs something else: a rough shared workspace, several people, movement interrupted mid-action rather than completed, and no glamorous isolation. Nobody hands anyone a key in a room built for two. Someone gets handed a key while already turning to answer somebody else.
Line those up and the conflict is not a matter of taste. It is specific and itemised. The reference proposes a locked-off two-hander in a controlled interior; the spot needs reactive framing in a crowded one. Those are different films, and the caption is not currently saying which one is being made.
Keep the distinctions honest here. "The room reads as staged" is a visible property — a reader can look and check. "This feels like the hotel bar where I met my wife" is a private association. Private associations are why the image is on the wall, and they are also the part that does not transfer to anyone else. When you list the hostile signals, list the ones that are in the frame, not the ones that are in your history with it.
And notice what this section is not claiming. It is not claiming that every collaborator will import every signal, and it is not claiming a universal law of image reading. It is an accounting exercise: the caption names one intention; the frame supplies material for many more. That gap is where the wrong film gets made, whether it happens this week or after three rounds of boards.
Write the caption twice
The first annotation is the one most treatments actually get:
Intimacy. Ignore the room.
This does almost nothing, in two ways. "Intimacy" has no stable referent in the frame — it is a word for the effect, pointing back at the thing it is supposed to explain. And "ignore the room" asks a reader to unsee the largest object in the picture. You cannot instruct someone past the dominant image. If they could do that reliably, the reference would not have been needed.
The second annotation points at an event:
Use: the pause before the hand accepts. The object is held a beat too long, then let go. Not: the room, the stillness, the styling, the two-person isolation.
This is better because it names something observable with a duration. You could measure the beat. You could shoot it. You could tell an actor to play it. The caption now has a job that a caption can do: it directs attention to one specific behaviour and flags the rest as incidental.
But watch what is still standing after the good caption. The room is still glossy. The camera is still still. There are still only two people, and neither of them is interrupted. The caption narrowed the reading; the frame still widened it. A caption is a paraphrase of a shot, and paraphrases lose.
Here is a workable test, and it is rough but it holds. Count what the annotation is disowning against what it is claiming. If the caption spends more words keeping the image away from the film than bringing it toward the film, you are no longer using the reference. You are litigating it, in public, in a document that other people have to work from. That is usually a sign to stop defending and start comparing.
Try a second source, and check whether it clarifies or averages
Pair the glossy reference with a spatial counterexample: a rough clip of a shared workspace, several people mid-task, one turning away from the frame, camera slightly late to the action, ugly overhead light, nobody styled. Nothing in it is beautiful. It is entirely about density, pace and interruption.
A pair can do something a caption cannot. It can split the questions. The first governs behaviour and timing — the half-second of hesitation. The second governs space and rhythm — how many bodies, how late the camera arrives, how much clutter survives the frame. Two references, two different questions, one coherent proposal. That is the pair working.
The pair fails when both references are answering the same question. If the first is your tone and the second is also your tone, you have handed the room a choice and called it a mood board. Whoever reads it will pick the one they preferred when they walked in. That is not a clarification; it is a delay with extra steps.
So before you add the second reference, say out loud what question each one answers. If they are the same question, you do not have a pair. You have a vote.
There is also a liability in pairing that people underestimate: a rough, unbeautiful reference loses arguments. The glossy one is seductive, and the workspace one is not. If the client or the creative director is drawn to the first, the second reference will get read as the compromise — the "realistic" option, the one that is being proposed because of budget. You may have introduced a correct description of the film and attached it to the losing side of an aesthetic argument you did not intend to start.
Compare replacement against keeping
The alternative to all this correcting is a reference that is less arresting and more accurate. Or no borrowed image at all.
An original drawing of the actual encounter will specify what a photograph cannot help specifying. A sketch has no grade, no lens character, no styling, no skin. It can put five people in the room and stage the moment of interruption exactly as the spot needs it. In one sense it is the cleanest instrument available: it carries almost no accidental signals.
It also carries almost none of the accidental gifts. The clip gave you a hesitation that someone actually performed, with an unplanned small delay that reads as real. A drawing has to build that deliberately, and if the drawing is rough — which is the point — it cannot hold the beat at all. You have traded a documented cue for clarity about everything except the thing you were trying to protect.
That trade is sometimes correct and should be made with eyes open. Two honest versions:
Keep with the specific caption. You retain the performed hesitation and accept that the room, the stillness and the finish will keep arriving in other people's work even after the caption names them. This is a deliberate, named risk rather than a hoped-for one. It is defensible if the behaviour is genuinely hard to describe and genuinely central.
Replace with a rough original. You get the space, the crowd, the interruption and the pace exactly right, and you lose the thing the borrowed image was doing. If you go this way, the hesitation has to be rebuilt somewhere: in the drawing's staging, or in prose next to it. Prose is precise and carries no texture, which is fine if what you need is an instruction rather than an atmosphere. Write the beat down: the object is offered; the hand opens; it does not close for a beat; then it closes. That sentence will not move anyone. It will also not be misread.
There is a middle route worth naming, because it is what most experienced people actually do. Move the emotional cue into language and let the image carry only what images carry well. Keep the drawn or photographed workspace for space and density. Put the delay in the caption as a behaviour, not as a mood. You lose the seduction and gain a document nobody can misread into a perfume commercial.
When the problem is a decision nobody has made
Sometimes the conflict is not in the reference at all. The brief says rough shared workspace and no glamorous isolation; somebody in the room is quietly hoping for the glossy two-hander anyway. Then the reference is being asked to settle a choice it did not cause and cannot conclude.
There are two versions of that, and they call for opposite moves.
In the first, the brief is genuinely open — nobody has decided whether this commercial is two people alone in a beautiful room or eight people in an ugly one, and both readings are live proposals rather than one plan and one mistake. Then stop editing the reference and return the question to whoever owns it. The reference cannot answer it, and every caption you write will be an attempt to smuggle an answer into the reference.
In the second, the brief has already answered it. The glossy hope is an unstated preference the brief rules out, not a competing proposal. No caption should be rewritten to accommodate it, and the reference is not what is standing in the way. Name the disagreement in the terms the brief already uses, and take it to whoever can hold the brief or change it. Reworking the reference to make room for a choice nobody will say out loud is how the wrong film gets made quietly, and the caption takes the blame.
Be careful with the language of diagnosis here, because the distinction matters. Without actual accounts from the people on the job, "the team will read this as luxury" is a prediction based on visible signals. It is not a finding. A real case would involve reading how the image is being used — which decisions it is generating, who has asked for what — and, if you want to quote anyone, asking them and getting permission. That work has not been done here, and the method above does not depend on pretending it has.
The three-way decision
Bring it back to the image on the wall.
Keep it when the useful cue is specific, performed, and hard to get any other way — and when you can name, in the same breath, exactly what the frame will keep saying that you don't mean. A kept reference with an honest caption is a real instrument. A kept reference with four paragraphs of disclaimer attached is a warning sign wearing a caption.
Pair it when the second reference answers a genuinely different question, and be ready for the fact that the plainer image will lose the aesthetic argument even when it wins the practical one.
Replace it when the correction has become the main content, and accept that you are giving up whatever the original did well. Then put that quality somewhere else — in staging, in a written instruction, in a second look at the frame.
Under all three, some misunderstanding remains. That is not a failure of the method; it is the condition of borrowing someone else's footage to describe a film that does not exist yet. The aim was never a perfect match. It is a reference whose misleading signals you have named before somebody else names them for you, and a decision you can defend without appealing to how the image makes you feel.
Keep the feeling when the filmmaking around it gets clearer — not when the defence of it gets longer.
Frequently asked questions
When is it worth keeping a reference that partly proposes the wrong film?
Keep it only when its useful cue survives the misunderstanding its other signals will introduce. Not when a disclaimer has been added to it, and not when the argument for keeping it has gotten longer than the description of what it contributes. The useful cue has to still be doing work after you subtract everything the reference says that you do not want. A kept reference with an honest caption can be a real instrument; a kept reference with four paragraphs of disclaimer attached is a warning sign wearing a caption.
How do I turn a feeling like intimacy into something a collaborator can build from?
Point at the moment in the clip that produces the feeling. In the constructed example, it is the pause before the hand accepts: the object is held a beat longer than a handover needs, then released. That moment is observable and has a duration, so it can be shot or played. If you cannot point at the moment that produces the feeling, you do not have a reference; you have a preference. A label like intimacy names a result and cannot be art-directed.
What makes a caption like 'Intimacy. Ignore the room.' weak?
It does almost nothing in two ways. Intimacy has no stable referent in the frame; it is a word for the effect pointing back at what it is supposed to explain. Ignore the room asks a reader to unsee the largest object in the picture, and you cannot instruct someone past the dominant image. A better caption points at an event, such as: use the pause before the hand accepts; the object is held a beat too long, then let go. Not the room, the stillness, the styling, the two-person isolation. Even then, the room is still glossy, the camera is still still, and there are still only two people, so some misunderstanding remains.
When does pairing references clarify, and when does it become a vote?
A pair works when the references answer genuinely different questions. One can govern behaviour and timing, such as the half-second of hesitation; the other can govern space and rhythm, such as how many bodies and how late the camera arrives. The pair fails when both references answer the same question: if the first is your tone and the second is also your tone, you have handed the room a choice and called it a mood board. Also, a rough, unbeautiful reference may lose the aesthetic argument even when it wins the practical one, and it can be read as the compromise or budget option.
What if the real conflict is a decision nobody has made?
Separate two situations. If the brief is genuinely open and both readings are live proposals, stop editing the reference and return the question to whoever owns it. If the brief has already answered it and the glossy hope is an unstated preference the brief rules out, no caption should be rewritten to accommodate it; name the disagreement in the terms the brief already uses. Reworking the reference to make room for a choice nobody will say out loud is how the wrong film gets made quietly. Be careful with diagnosis: without actual accounts from the people on the job, a claim like the team will read this as luxury is a prediction based on visible signals, not a finding.