Pitch an Audio-Only Commercial Without Describing a Film Nobody Will See
Pitch an Audio-Only Commercial Without Describing a Film Nobody Will See
The trouble starts in the first ten seconds. A creak happens, and the listener has no way to know what creaked. In the board the client approved, a hand was resting on the chair back before the noise arrived, so the object was identified before it spoke. In the audio version the hand is gone. What's left is a sound, an unplaced voice, and a growing suspicion that this is a commercial about a chair, which nobody has said yet.
The answer is not to narrate the missing footage. It is to rebuild the listener's route to the facts. For every thing they have to end up knowing—who is speaking, what is making the noise, what changed, whose offer this is—decide which audible event carries it: a line, a sound, or a silence. If two events carry the same fact, one of them is decoration. If none carries it, the listener will fill the gap with whatever the sound reminds them of, and the ad becomes a description of a commercial that isn't playing.
List what the pictures were explaining
Before writing a line, make an inventory of facts rather than shots. Take the visual version and ask what it teaches the audience, and which of those lessons have already survived into sound.
A creak tells you that something made a noise. It does not tell you that the something is a chair. A chair's legs scraping a floor tell you that furniture moved, but not which furniture, or whether it is being pulled out, pushed back, or dragged across a room. Spoken dialogue tells you a lot—if the speakers already have a reason to say it. That is the trade.
Four kinds of fact are worth separating, because they fail differently.
The relationship. Who are these two people to each other? In the visual version you know within a second that one lives here and the other is visiting, because she is standing and he is waiting to be told where to sit. Nothing in the audio says that unless one of them behaves like it or mentions it.
The object in use. The chair. The picture identified it with a hand on the back; the sound alone identifies only an event.
The change. What is different at the end of the spot? If the answer is "he now knows the chairs have a problem and she has a shop for it," then the ending has to be audible. If the answer is "the chair got fixed," the whole plan has a problem, which is one reason to write the offer rather than the outcome. In the version developed here, nobody fixes anything: the scene is the problem, the shop is named, and no repair is claimed or shown in either medium.
The brand's position. Which business is speaking, and what does it do? This is the one fact with no substitute. A logo card, an end frame, a van in the background—all of those are carriers in a different medium. In audio, the business has to be identifiable in the audio. Spotify's own guidance for audio ads is explicit on that point, and lists voiceover, music and timing as production elements (page recorded as checked 8 September 2026). That is a publisher describing its own format, so take it as a floor, not a law. The example ads on that page weren't played for this piece, so nothing here rests on how they sound or how long they run.
The point of the inventory is diagnosis, not a script. Half the items on it are mood rather than fact. Whether the room is a kitchen or a dining room, whether it's morning, whether anyone is amused—those can stay unfixed if the spot doesn't turn on them. What matters is finding the two or three facts the ad collapses without, and noticing that none of them currently has a carrier.
Choose the voice that can carry the scene
There are three broad routes, and they are not interchangeable insurance.
Dialogue builds the relationship out of what people want from each other. Tom asks where to sit, tests the chairs, complains, and asks how bad it has to get before somebody is called. Each of those is a small act, and the acts add up to a relationship: he is a guest, she is at home in this, she is not embarrassed. Dialogue is bad at announcing what the room already shows. Nobody in a kitchen says that the table has four chairs. It can carry a count once a character has a reason to arrive at one: Tom can say "two of your four" because he has just tested all four and has something to report, and Nadia can answer "my four" because his "your" gives her a reason to claim them. Counts reach dialogue as discoveries and replies, not as standing facts.
Narration can supply the standing facts instead—four chairs at one table, whose they are, how long the problem has gone unfixed—without needing anyone to have a reason to say them. The cost is pressure. A narrator who explains the setup usually keeps explaining, and the ad becomes a description with sound effects.
A sound-led opening puts the creak and the scrape first and lets the voices arrive second. It can establish that a room is occupied and something is being moved, but on its own it cannot turn an unidentified noise into a chair. At some point a line has to convert the sound into a fact.
Here is the same scene carried two ways. Four chairs, one table, two of them noisy. Nadia's chairs; Tom visiting. The shop is Marlow & Daughters, and it works on chairs.
Dialogue-led:
TOM: This one?
NADIA: Any one you like.
[A chair creaks, long and rising.]
TOM: Is that the chair, or the floor?
NADIA: That's the chair. That's the musical seat.
[Scrape of chair legs. Room tone.]
TOM: And this one's shy.
[Scrape.]
[Scrape. A creak, shorter.]
TOM: — and this one sings too. So two of your four.
NADIA: Two of my four.
TOM: How many before you call somebody?
NADIA: Marlow & Daughters. They do chairs.
Narrator-led:
NARRATOR: Four chairs at one table. Only two of them keep quiet.
NADIA: Sit anywhere.
[A chair creaks, long and rising.]
NARRATOR: That's one of the noisy ones.
NADIA: That's the musical seat.
[Scrape. Room tone.]
NARRATOR: The next one has nothing to say.
[Scrape. A creak.]
NARRATOR: That's the other one.
TOM: So two of your four.
NARRATOR: She has been meaning to call somebody about it for a while.
NADIA: Two of my four. Marlow & Daughters. They do chairs.
The dialogue draft spends one line to do what the picture did for free. "Is that the chair, or the floor?" is not elegant, but it is plausible, and it converts a noise into an object at the moment the listener needs the conversion. The narrator draft buys the table and the count in its first line and the ownership a few lines later, then keeps spending. By its fourth line it is saying things the creak already said.
You could combine them. That is the instinct, and it usually costs the scene, because two organising voices means every fact gets told twice—once by the room and once by the narrator—and the listener starts waiting for the narrator instead of listening to the room. Pick the carrier that fails least, and let the other facts ride on inference.
Give each sound a job the listener can recognize
Sounds in a treatment fall into a few rough jobs, and only one of them has to be understood the same way by every listener.
Atmosphere sets the room. Room tone, traffic, weather, the hum of a fridge. Nobody has to identify these correctly; they only have to make the scene habitable.
Action cues mark something moving. The chair-legs scrape, a door, a kettle, a hand on a table. These can be read loosely. A scrape says "something was moved with weight behind it," and that is enough when a line is holding the object.
Transitions mark a change of place or time. A door closing, a car pulling away, a phone going quiet. Useful for leaving one scene for another, and rarely load-bearing on its own.
Premise information is the dangerous one. Here the sound is the fact—the creak is the whole reason the ad exists. A cue carrying a premise fact has to land unambiguously, because the ad doesn't work if the listener decides the noise was a hinge, a swing, or a shopping trolley.
In the dialogue-led draft, the scrape is doing the action job and the line is doing the object job. That division is worth writing down in the treatment, in exactly those terms, because it tells the audio director which cues are allowed to be impressionistic and which ones are not. "The scrape establishes movement; the line establishes the object" is a useful sentence. "Replace the door slam with a lighter thud" is a mixing note that belongs in a later conversation.
Silence deserves the same listing. In the draft above, the quiet chair is carried by a beat with nothing in it but room tone and the scrape. That only works because the first creak has already been attached to a chair. Silence can confirm a fact the listener holds; it cannot supply one they don't have.
Two cautions for the treatment document itself. First, don't stack cues on the same fact. If the chair creaks, the legs scrape, and somebody says "hear that chair," the listener is being told three times and the scene sounds nervous. Second, describe the relationship between voice, sound and silence rather than the settings that will realise it. A treatment can say the creak sits at the volume of a real chair in a quiet room and the voices stay clearly on top of it. It should not say how. It also shouldn't promise a particular track or a licensing outcome that hasn't been cleared—say what the scene wants from music, if anything, and mark the rights question as open.
Order the information, then hear the draft
Listeners can hold about one new thing at a time, so order does as much work as content. The creak comes before the naming. If Nadia says "the musical seat" first, the creak becomes the thing everyone was waiting for, and her line turns into a caption. Sound first, name second, reaction third is the sequence that lets the naming land as a naming.
The brand has a fixed position: somewhere in the audible material, the business and what it does. "Marlow & Daughters. They do chairs" does that job without claiming that any chair was repaired, which is a distinction worth protecting. The offer is the shop's existence and its category; the result is a different story the ad isn't telling.
Then there is the read-through, which is a different exercise from proofreading. Read the lines aloud as they'd be performed, with the voices the scene implies, and let the pauses take real time. Then ask someone who hasn't read the treatment and hasn't seen the board to say what happened. Ask specific questions rather than "did you get it": who were those people, what was making the noise, what does the business do. Write down their phrasing, not your impression of it. Their summary is the actual script; the one on the page is a hypothesis.
Two things not to do. Don't infer duration from word count—a line with a pause and a scrape around it is not the same length as the same line read flat, and any seconds quoted in the treatment before an actual read are placeholders. And when something is missing from the listener's account, don't fix it by reading faster. Fix the missing information.
None of that reading has been done here. The two drafts above are proposals; no recording exists, no listener has been tested, and no duration is claimed.
The sequence, with a carrier for each fact
Assembled, with the table question left where it belongs:
NADIA: Sit anywhere. End one's my favourite.
[A chair creaks, long and rising.]
TOM: Is that the chair, or the floor?
NADIA: That's the chair. That's the musical seat.
[Scrape of chair legs. Room tone. Nothing creaks.]
TOM: This one's shy.
[Scrape. A creak, shorter.]
TOM: — and this one sings too. So two of your four.
NADIA: Two of my four. I keep meaning to call somebody.
TOM: Somebody in particular?
NADIA: Marlow & Daughters. They do chairs.
Check it fact by fact. Something creaked, and the chair is named within a line, so the object holds. The second noisy chair gets its own shorter creak rather than leaning on Tom's adjective, so "two of your four" is a count the listener has heard, not just been told. There are four seats, carried by "two of your four." The seats are Nadia's, carried by "my four." The problem is chronic and unresolved, carried by "I keep meaning to call somebody." The host's fondness for the noise—the thing that makes her insist on "the musical seat" instead of apologising—is carried by "End one's my favourite" and by the naming itself. The shop's name and category are spoken. No repair happens.
What's still missing is the table. The listener knows there are four of something. A row of chairs against a wall would sound exactly like this. That's a real gap, and the honest responses are limited: spend four words on it, hand the opening line to a narrator, or decide the table was never doing any work for an ad about chairs. Any of those is a choice about this spot. What none of them is, is a reason to reach for a paragraph of voiceover every time an audio adaptation loses a picture. The missing table was exposed by reading the sequence and counting what it actually told someone. That's a cheaper habit than describing a film nobody will see.
Frequently asked questions
How do you adapt a visual commercial to audio without narrating the missing footage?
Make an inventory of facts rather than shots, then decide which audible event carries each fact: a line, a sound, or a silence. If two events carry the same fact, one is decoration. If none carries it, the listener will fill the gap with whatever the sound reminds them of, and the ad becomes a description of a commercial that is not playing.
Which facts are hardest to carry in audio?
The relationship, the object in use, the change, and the brand's position. Relationship must be shown by behavior or mentioned; an object needs a line to convert noise into a thing; the change has to be audible; and the business must be identifiable in the audio itself, because a logo card or end frame is a carrier in a different medium.
When should audio use dialogue, narration, or a sound-led opening?
Dialogue builds the relationship from what people want from each other and can carry counts as discoveries and replies, but it is bad at announcing standing facts. Narration can supply standing facts, but it tends to keep explaining and turn the ad into a description with sound effects. A sound-led opening can establish an occupied room and movement, but on its own it cannot turn an unidentified noise into a chair; a line has to convert the sound into a fact.
What jobs can sounds have in an audio treatment?
Atmosphere makes the room habitable; action cues mark movement and can be read loosely; transitions mark a change of place or time; and premise information is the dangerous one because the sound itself is the fact and must land unambiguously. Silence can confirm a fact the listener already holds, but it cannot supply one they do not have.
How should an audio draft be tested?
Read the lines aloud as they would be performed, with real pauses, then ask someone who has not read the treatment or seen the visual board to say what happened. Ask specific questions: who were those people, what was making the noise, what does the business do. Write down their phrasing. Do not infer duration from word count, and when information is missing, fix the missing information rather than reading faster.