Make a Sound-Only Sketch for a Commercial Pitch

A pitch meeting can absorb an hour of sound adjectives and still not know what the spot sounds like. Warm but not soft. Sparse, with something underneath it. It builds. Everyone nods; nobody has agreed to anything, and the argument resumes next week with a slightly different set of adjectives.

A playlist won't settle it. A reference track tells you what a sound is like. It doesn't tell you what happens when. Most of the meaning in a short piece of sound sits in order — whether the jingle arrives before the drawer opens or after, whether a voice lands on the movement or a beat behind it. Adjectives can't hold that. A short arrangement can.

That is the whole job of a sound-only sketch: put one listening idea into time, in a form you can still move. Not a mix, not a soundtrack, not a general argument that sound matters. A provisional arrangement, built from material you're allowed to use, that lets someone hear a specific proposal and then disagree with it.

Choose the listening question

Start by naming one question you could say out loud in a sentence. "I want to know whether the keys are heard before the drawer opens." "I want to know whether the voice should sit on top of the movement or behind it." "I want to know whether there's room for a line at all, or whether the room tone is eating it."

A collection of attractive tracks is not yet an audible proposition, because it doesn't commit to anything. The test of a good question is whether you can build two versions that differ only in that one relationship, and whether someone who heard both could tell you which they preferred and roughly why. If the honest answer is "it depends on the picture," you've picked a question a sound-only sketch can't answer, and you should choose a different one. A sketch that answers ten questions is a mix, and a mix is a much more expensive thing to be wrong about.

Find the moment the sound is supposed to explain: an approach, a withheld reveal, a shift in where attention sits. Then write the question down somewhere you'll see it, because it will try to expand while you work.

Build separate tracks from authorized material

Separate tracks exist for one reason: so you can move one element without moving everything else. Voice, significant effects, and ambience each get their own track. Nothing else needs one until you find a clip that has to be nudged on its own.

The example I'll use for the rest of this piece is a plan, not a recording. Nothing in it has been captured, arranged, or listened to. It's a twelve-second scene: someone hunting a key ring in a drawer. Four elements, four tracks.

• VO — line. A short spoken phrase, recorded close and dry. Your own voice is the simplest option. One line, no performance.
• FX — drawer. The handle and the slide, recorded near the actual drawer.
• FX — keys. The ring being stirred, then lifted.
• AMB — room. A minute of the room doing nothing, which is all you need for a twelve-second bed.

Name them in a way that survives a week. "FX — keys" beats "keys2final." Someone else will open this session, possibly you.

Physical recording beats library shopping here, and not for purity reasons. A downloaded jingle and a downloaded drawer sound won't tell you whether your timing works; they'll tell you whether someone else's recording is convincing. Keep the raw takes untouched in their own folder and work on copies. The raw files are the only part of this you can't reconstruct.

Then there's music. If you're tempted to drop in a track to convey the mood, notice that you've just changed the artifact: you're no longer testing an arrangement, you're testing someone else's composition underneath it. A temporary reference inside a private session is one thing. A file you send out is another, and the permission question arrived the moment the file left the room. Keep borrowed music out unless its use is separately cleared.

Place entrances, overlaps, and deliberate gaps

Now build the sequence around the question, and change one relationship at a time.

Version A puts the keys first. Version B withholds them until the drawer has opened, and holds a gap before the lift. Same four sources, same total length, same export settings. Only the timing moves — and in B it moves in two places, not one.

Both versions run twelve seconds, which matters more than it sounds like it should. If one is nine seconds and the other is twelve, the person comparing them is partly reacting to length rather than to the key placement, and you'll never get that confound out of the conversation.

In version A, the early stir comes from inside a closed drawer, so it has to be smaller and duller than the lift, or the listener will place the keys out on the table and the question dissolves. That's the kind of detail the sketch exists to expose: the plausibility of the idea is audible, not theoretical.

Watch what your edits cover up. Moving the drawer earlier may now sit under the line, or the drawer slide may swallow the lift. Check the spoken line against the ambience by ear in each version. If the line is being covered, you have a real finding — but only if the line was meant to be heard.

Silence is an entrance too. The 1.7-second gap in version B is not empty; it's a held breath before the same jingle that version A plays earlier. That gap is why version B makes two moves at once, and you should be honest with yourself about that. If the room prefers B, you won't know whether the delay did the work or the gap did. So build a third version, B2: keys at 0:05.9, right after the drawer finishes, with no gap at all. B against B2 isolates the silence. That's the whole discipline — one relationship per comparison, and a new test whenever you can't attribute the effect.

Listen to the combined sketch without destroying it

Playback mixes the tracks so you hear them together. That's your listening situation, and it leaves everything intact. Export also mixes, but writes the result to a file instead of playing it. Alongside those, an explicit mix or render command produces a combined track — and the Audacity manual's mixing page distinguishes doing that into a new track from doing it by replacing the originals with the mix. Which one you get depends on the command you pick, so pick deliberately. The combined track is a convenience; the separated tracks are the sketch. Don't end up with only the convenience.

One warning worth thirty seconds of your time. Muted material is not reliably absent from every explicit mix operation. Mute one track, run the exact command you intend to use, and listen to what comes out. If you skip that test, you may hand over a file containing a layer you thought you'd silenced — or lose one you meant to keep — and you'll discover it in front of other people.

The manual page I'm leaning on here describes itself as development documentation, which is a fair heads-up that the menus and labels may not match the release installed on your machine. Check the commands in the version in front of you before treating any of this as a description of it.

Listen to the files back to back, not three days apart, and change nothing in between. There are three now — A, B and B2 — and two clean comparisons inside that set. A against B2 isolates keys-lead against keys-follow: B2 withholds the keys until the drawer is done and opens no gap, so their ordering is the only thing left in the difference. B against B2 isolates the held stretch: both let the keys follow the drawer, and the gap is all that separates them. Keep each pair matched in length and export settings.

Listen for three things in those comparisons: whether the line survives, whether the key action gets buried under the drawer, and whether the gap reads as a held moment or as a dropout. A gap with room tone still running underneath reads as held breath. A gap with absolute nothing underneath reads as a mistake, because that's what it sounds like when a file is broken.

Export and check the actual listening file

Save the session before you export anything, and save it again under a name that says what it is. The session is the artifact that lets someone else change one entrance instead of rebuilding the whole scene.

Export the provisional mix, then reopen the exported file and listen to it outside the session. That file is what the person on the other end actually receives; hearing it only through your own session is hearing a different thing. Check four facts: the right version is in it, the right sources are in it, both versions are the same length, and the muted tracks did what you believed they did. Keep the export settings identical across versions so the comparison isn't quietly measuring the container instead of the timing.

Then label the file in the file itself — in the metadata, in a companion note, or in the name. Something like keys-lead_vA and keys-follow_vB, plus a line stating what question they answer, that all sounds are your own recordings, that no licensed music is present, and that these are sketches rather than finished mixes.

Which brings up the last thing to leave off. Don't attach broadcast delivery specifications, loudness targets, or platform requirements to a sketch. Those belong to a later job with its own measurements and its own person responsible for them. An exported file is not evidence that the idea works; it's evidence that the idea is now audible.

What the room gets back

Three things go back: the editable session, two or three labeled listening files, and one answerable question. The question should be short enough to answer in a sentence — keys before or keys after; the gap or no gap.

If the answer comes back "before," you've settled a timing decision that survives into the shot, and you've settled it for the price of an afternoon. If the answer comes back "neither," you've learned your question was wrong, which is cheap to find out now and expensive to find out on the day. Neither outcome requires the sketch to be good. It only requires the sketch to be movable, so the next version costs one clip rather than the whole scene.

The sketch isn't a rough draft of the spot. It's a question with a duration, and its final value is that anyone can move the pieces.

Frequently asked questions

What is a sound-only sketch, and how does it differ from a playlist or reference track?

A sound-only sketch puts one listening idea into time in a form you can still move. A playlist or reference track tells you what a sound is like, not what happens when. Much of the meaning in short sound sits in order, such as whether the jingle arrives before or after the drawer opens. The sketch is a provisional arrangement built from authorized material, not a mix, soundtrack, or general argument that sound matters.

How do you choose the listening question for a sound-only sketch?

Name one question you could say out loud in a sentence, such as whether the keys are heard before the drawer opens. The test is whether you can build two versions that differ only in that relationship, and whether someone who heard both could say which they preferred and roughly why. If the honest answer is that it depends on the picture, choose a different question. A sketch that answers ten questions is a mix, and a mix is more expensive to be wrong about.

How should tracks and source material be set up?

Separate tracks exist so you can move one element without moving everything else. Voice, significant effects, and ambience each get their own track. Use material you are allowed to use; physical recording beats library shopping because downloaded sounds will not tell you whether your timing works. Keep raw takes untouched in their own folder and work on copies. Keep borrowed music out unless its use is separately cleared.

How do you isolate timing relationships and deliberate gaps?

Change one relationship at a time. For example, compare a version where keys lead with a version where keys follow the drawer. If the following version also adds a held gap, build a third version with the keys right after the drawer and no gap, so the silence is isolated. Keep versions the same length and use identical export settings, so the comparison is not quietly measuring length or container instead of timing. A gap with room tone still underneath reads as held breath; absolute nothing reads as a mistake.

What should be exported and handed over at the end?

Save the session before exporting, and save it again under a name that says what it is. Export the provisional mix, then reopen the exported file and listen outside the session. Check that the right version is in it, the right sources are in it, both versions are the same length, and muted tracks did what you believed. Label the files with what question they answer, state that all sounds are your own recordings and no licensed music is present, and call them sketches rather than finished mixes. Hand over the editable session, two or three labeled listening files, and one answerable question. Do not attach broadcast delivery specifications, loudness targets, or platform requirements to a sketch.

Further reading

• Record a Scratch Voice Track That Lets You Judge the Commercial’s Timing — Before you record the sketch's spoken line, this shows how to capture a scratch take clean enough to trust and provisional enough to throw away.
• Pitch an Audio-Only Commercial Without Describing a Film Nobody Will See — If the sketch may become an audio-only spot, this explains how to rebuild the listener's route to facts the picture used to carry.
• Write a Sound and Music Approach That Changes How the Commercial Plays — If the pitch brief still speaks in mood phrases, this shows how to turn them into audible actions the sketch can actually test.