How to brief sound for an ad half its audience will mute

01Do people actually watch ads with the sound off?

Enough of them that you have to build for it, and I am not going to tell you the share, because I cannot stand behind a number for it. Design for two rooms and the number stops mattering.

The behavior underneath it is not really about preference either. It is about context. A phone in a hand on a train is muted by circumstance, and the same person with the same ad in bed at eleven at night is not. You are not designing for a segment. You are designing for one person in two situations, which is why the answer is never to pick a side.

02Captions are not the answer to a sound problem

Captions solve exactly one thing: the words. They are mandatory on anything with a voice, and they are not a mute strategy. A film whose captions carry the entire script does work with the sound off. A film whose captions are sparse chapter headings does not, and the difference is binary rather than gradual. We have watched a film fragment its script into two and three word beats and become unwatchable in silence.

There is also a floor underneath them that people skip. Every line has to hold long enough to read: at least 1.2 seconds, or 0.35 seconds per word if that is longer, at full opacity, with fades not counting because text at partial alpha is not readable. And it has to survive being resampled to feed width. That whole argument lives in the legibility floors a caption has to clear on a phone.

03What carries meaning with the sound off?

Five things, and they are all decisions you make before a bed exists. Cut rhythm, which is the film's pulse whether or not anything is playing. Type, which is the only channel that can state a claim. Motion, meaning something changing on screen rather than a camera move over a still. Faces, which do more emotional work than any other picture element. And product legibility, which is whether a stranger can tell what the thing is.

The rule that falls out of this is short and we enforce it at the paper stage: nothing in the film's organizing idea may exist only in the audio. Sound is allowed to make a thing land harder. It is not allowed to be the thing. Build the idea in picture and let the sound sharpen it, and you have one film instead of two.

If the ad only lands with sound, it does not land.

Our own builder brief, where muted is a scored dimension rather than a caveat

04Write a mode plan, not a music note

A music note says "upbeat, modern, not too corporate." A mode plan says what the ears get in each section of the film and at which second it changes. It is a direction decision, made on paper alongside the beats, because changing register in a finished cut means re-cutting the picture around it.

Decide the register with it. Speech with gaps is a creator talking head, scripted voiceover with no gaps is a brand film, music with no speech is a pose reel. That separator is audible with your eyes shut, and it sets the cut rate and whether a bed is allowed at all.

Calling this a direction decision is not a formality. A mode plan changes the picture. If the first two and a half seconds belong to a voice, the opening shot has to hold long enough for a sentence, which is a different shot from the one a music-led opening wants. Decide it after the cut exists and you are choosing between re-editing and settling. We have settled, and the settling is visible in the film. It gets drawn on the same page as the picture in how a short ad is built.

05How do you brief a track you have not heard yet?

By its job and its events, not its genre. A genre label buys you a mood and no timing. What a composer or a generator can actually act on is: where the first sound is, what the arc shape is, which single moment gets the sharpest thing in the mix, where the mix has to get out of the way, and how it ends. Here is the whole page.

SOUND BRIEF - one film, one page
film        <name>   runtime 15.0s   24 fps   9:16

1  MODE PLAN            what the ears get, section by section
   0.0 - 2.5    voiceover, dry, no bed underneath
   2.5 - 11.0   music body, bass-led
   11.0 - 15.0  room tone close, music out on the cut at 11.0
   register     scripted voice, gaps removed. this is a brand film

2  POLE                 bass-led. nothing sits in the middle

3  FIRST SOUND          frame 0. a transient, never a swell
   source       the clasp closing. dry, no room, about 120 ms

4  THE ONE SHARP THING
   6.4s         the stream hitting ceramic
   rule         one high-frequency event. it appears nowhere else

5  ACCENTS              all computed from the cut table, never fixed frames
   6.4s         sharp event, 9 frames after the cut, inside the beat
   7.6 - 8.1    engineered gap. picture moves before the sound returns
   8.1s         drop, inside the re-hook beat, never on its first frame

6  THE YIELD            mix steps back while the viewer is reading
   4.2 - 5.6    line two on screen. sub down, mid up
   11.0 - 12.4  lockup. mix reclaims at 12.4

7  ENDING               last 2s at >= 0.60 of the film's peak 2s level
                        no tail-out, no quiet card

8  MUTED CONTRACT       the film this becomes when nobody presses unmute
   proposition legible before frame 12
   every line holds max(1.2s, 0.35s per word), fades excluded
   cut rhythm carries the argument with no help
   nothing in the organizing idea exists only in the audio

9  DELIVERY             stems: voice / bed / design, plus a mixed master
                        tempo declared, then read back off the render

10 WHAT I DO NOT WANT
   a swell into frame 0
   treble spread through the bed instead of on the one sharp thing
   a bed reused from another film in this campaign

Two lines in there do more work than the rest. Section 5, because hits computed from the cut table cannot drift when a shot length changes, and hits nailed to frame numbers always do. And section 10, because a list of refusals is the fastest way to save a round: it is easier to say what you will reject than to describe what you want, and this is the same reason a reference video works better than a brief.

06How do you audition a bed in thirty seconds?

Against the cut, on a phone speaker, twice. Never on its own and never on headphones. A track auditioned alone is being judged as music, and you are not buying music, you are buying the second half of a film that already exists.

07What to send a composer, and what to send a generator

The same brief, weighted differently. A composer can act on intent and will ask better questions than the document contains. A generator cannot ask anything, so the parts a person would have queried have to be decided in advance and written down.

The reason this is a briefing problem and not a mixing problem is that everything expensive about it is decided before anybody opens a session. A film that needs its voiceover to make sense was doomed at the story stage, not at the mix. If your ad only works with the sound on, the sound is not what is wrong with it. The rest of the ordering argument is in sound is 40% of the ad and in how to brief and QA ad creative.