Design the Silent Version Without Turning Every Sound Into a Caption

The brief asks for a version that plays with the sound off. It doesn't ask you to mute the timeline and sprinkle text wherever something goes missing. Those are different jobs, and only one of them produces something a viewer can follow.

Muting tells you what breaks. It doesn't tell you what to build in its place. The rebuild is usually smaller than people expect — one visible cause, one re-blocked action, one line of text placed where it has room — but it has to be the right one. Label every sound instead and you get a transcript laid over an edit that was never cut to be read.

Three things the soundtrack was actually doing

Before you touch anything, go beat by beat and sort what the audio is carrying. It's rarely one thing.

Speech that supplies a fact. Someone says the thing the story turns on. Lose it and the audience is missing information, not atmosphere.

Sound that identifies a cause you cannot see. A chime, a knock, a door. The reaction in frame tells you something happened; the sound tells you what. This is the category people undercount, because the visual — a face, a flinch, a head turning — feels like it's carrying the moment. It isn't. A startled face says something happened. It does not say whether it was a doorbell, an alarm, or a dropped glass.

Music that changes how a retained action reads. This one gets missed entirely because nothing disappears from the picture. Take the music away and the same shot can mean something else. Not less — else.

That third category is why "is the story still recoverable?" is the wrong test. A muted spot often still makes sense in retrospect. The loss is in the timing of the inference, and in what the audience guesses while they're waiting.

An invented scene, so the arithmetic stays checkable

The following is a constructed example, not a production. Nothing here was shot, tested, or shown to anyone. Its numbers are stipulated so the comparison holds still.

A 20-second spot, 16:9. One desk, one worker, one small ribboned box. A phone sits on the desk within frame but at an angle, its screen turned so the text isn't legible. A door sits at the back of frame with a narrow glazed panel; the corridor beyond is brighter than the room.

Note what's already working. The phone lighting up is a visible cause. The viewer sees that something arrived on that screen. The problem is that the cause is visible and the content isn't — so the audience knows an event happened and has to guess what it was.

They will guess. That's the part worth internalizing. An under-specified cause isn't neutral; it's a slot the viewer fills with the most familiar nearby convention. A phone lighting up in an office and a person hurriedly hiding something reads, by default, as trouble. Workplace problem. Evidence. Bad news. It does not default to birthday present.

Mute it and watch the clock

Run the same 20 seconds with no audio and no captions.

At 0:04 the screen lights and the worker reacts. The viewer sees a cause but can't read it. At 0:06 the box goes into the drawer — an action performed for no stated reason. At 0:11 the worker turns to the door, which reads as idle looking. At 0:14 the door opens and a visitor arrives.

Now the audience can reconstruct it: someone came in, so the worker hid the box from them. The story is recoverable. It's just recoverable late — the drawer is shut by 0:11, and the reason for it lands at 0:14, three seconds after the hiding is over and ten seconds after the screen lit. In the meantime the hiding has been filed under the wrong genre.

That gap is the whole problem. Someone is coming doesn't land until the door opens, which is after the hiding is finished. Why the box moves lands only in retrospect, if it lands at all. The spot isn't broken. It's just no longer in on its own joke.

Four ways to fix it, ordered by cost

The instinct is to reach for text first. Two different things get called text here, and the ranking below covers only one of them. A message the audience reads inside the shot — the phone's screen in this scene — is a photographed object; making it legible is a change of framing, not a new layer, and it is the first entry below. A line or graphic laid over the picture is a layer. It can carry the fact, but it spends a reading clock the original cut never paid for, and it comes after these four rather than before them; that cost gets its own section.

1. Give the cause a visible source. The phone is already there and already lights up. Turn it toward camera at the top of the scene, or push in enough that the message is legible at 0:04 and 0:05. Nothing new enters the frame; the existing cause simply starts saying what it means. This restores both facts in time: someone is coming at 0:05, why the box moves at 0:06.

The cost is real. A lit screen at readable size competes with the worker's face, and legibility in a 16:9 frame is not legibility in a vertical crop, where the phone may be smaller, moved, or out of frame entirely. Check the delivered shapes before you commit.

2. Move the cause to the other end of the room. Drop the notification from the picture altogether and let the arrival announce itself. The corridor behind the glass panel is brighter than the room, so a dark shape crossing it reads. The worker catches it in peripheral vision, half-turns, and hides the box. Same clock: someone is approaching at 0:05, why the box moves at 0:06.

Two costs. The door and its panel have to stay in frame and stay lit for the silhouette to read at all, which is a set and lighting dependency rather than a copy change. And this doesn't restore the notification — it deletes the need for it. "Someone is outside" replaces "your visitor has arrived." If the story only needs someone is about to walk in, that's a fair trade. If it needs the person the gift is for, the silhouette never says it.

3. Change the action so the intent is visible. The worker's performance is doing unpaid work in the muted cut. In the scored version, the pizzicato made the hiding playful. Without it, the same sweep into the drawer reads as concealment with a bad reason. A grin on the way down buys back most of that reading, and it's a performance note, not a graphics note.

There's a version of this where the loss isn't worth repairing. Muting doesn't only subtract information; it re-reads actions you kept. Ask whether the new reading breaks the proposition. If the spot is about a surprise someone is excited to give, a guilty read is fatal. If the spot is about someone having a secret, the same read is fine, and you've saved yourself a rewrite.

4. Accept that the idea needs a different version. Sometimes the sound is the proposition — the joke is the ringtone, the turn is the track dropping. No visible cause restores that, because the thing being lost isn't information. It's the premise. Build a different 20 seconds around the same character and situation rather than pasting labels onto the old one. This is a legitimate outcome, not a failure, and it's cheaper than forcing the original to survive without its engine.

Reading and watching need separate moments

Whichever kind of text you choose — a line laid over the picture or a screen made legible inside it — place it where nothing else is competing for the same attention.

The temptation is to put the words exactly where the sound used to be, because that's where the beat sits. But the original beat was cut to audio, and audio doesn't have a reading clock. A two-second spoken line becomes text that takes longer than two seconds to read, and it runs past the beat it was meant to explain. Then the drawer action happens underneath text the viewer is still finishing, and one of the two is lost.

Two fixes. Move the text earlier so its reading finishes before the action starts, or cut the information down until it clears inside the beat. What you shouldn't do is shrink the type until it fits. Shrinking type doesn't create reading time; it just makes the text slower to read at the size it will actually be seen.

Do the arithmetic against your own delivery — frame shape, duration, where the screen will be. There's no universal number for how many words fit in a beat, and a line that works at arm's length in a widescreen frame is a different proposition in a vertical crop on a phone. The only honest test is to watch the actual version, at the actual size, with the actual duration.

If the campaign also delivers a shorter cut, check that the reading survives the compression too. A visual cause that is legible in 20 seconds may not be legible in six.

The captions are a different deliverable

The version you just redesigned and the captions for the original are not the same object, and treating one as the other causes most of the confusion here.

W3C's Web Accessibility Initiative guidance on captions, which carries a September 2024 update, describes captions as synchronized text for speech and for non-speech audio information that matters to understanding, and distinguishes the term from subtitles. That definition bounds the distinction; it doesn't validate a creative redesign, and it doesn't certify any particular video as accessible. Nothing in this article does either.

What the distinction means in practice: captions are built on the assumption that the audio exists. They describe it. So a caption reading something equivalent to [phone chimes] is accurate for the version that has a chime — and it's a small oddity in a version that has no sound at all, telling the audience about an audio event they aren't experiencing. It also doesn't fix the cause. A label that says a sound happened is not the same as showing what made the sound.

There's a mechanical reason captions alone make a poor sound-off version: they were timed to the audio they describe. Correct them and they will still be a text layer over an edit that was cut for the ear. That's not a flaw in the captioning. It's a mismatch of jobs.

Two practical consequences:

Captions have to be turned on. The sound-off version can't depend on a setting, a platform behavior, or a menu the viewer never opens.

Don't delete the sound from the master to force a solution. The captioning work needs the original mix to describe it. The sound-off version is an additional deliverable alongside the captions, not a replacement for them.

Review it by asking what happened

You are the worst possible viewer of your own fix, because you already know the answer. Watching it once and deciding it reads clearly tells you almost nothing.

Get someone who hasn't seen it. Mute the version, play it once, and ask them to describe what happened — not whether it felt clear. "Does it read?" invites politeness. "What happened, and why did she put the box in the drawer?" produces a description you can hold against the story you meant to tell. Listen for which facts they volunteer and, more importantly, for where they put them in the sequence. A reviewer who says she hid a present because someone was coming got everything you needed, in the right order. A reviewer who says she hid something and then someone walked in got the events and missed the causality.

If a reviewer supplies a cause you didn't intend — bad news, a warning, an affair — that's the under-specified slot filling itself in. Record it and revise the example rather than dismissing it. And keep the accessibility review separate: a caption check asks whether the text accurately conveys the speech and meaningful audio of the original. It answers a different question from whether the silent version tells the story.

What the redesign should leave behind

For this invented spot, the smallest change that restores both facts and their timing is making the phone legible — same object, same beat, cause already in frame, no new insert. Someone is coming at 0:05, why the box moves at 0:06, and the audience gets to enjoy the near-miss instead of reconstructing it afterward.

Two conditions attach. The phone has to stay legible in every delivered frame shape, and the screen can't steal the beat from the worker's face. If either fails, move the cause to the door and let the silhouette carry it, accepting that someone is outside replaces your visitor has arrived.

The third change doesn't have a clean fix and shouldn't get a fake one. The music made the hiding playful; silence makes it furtive. Pay for that in performance with a grin, or decide the colder reading is acceptable, or conclude that the idea needs a different version.

What's left is the review — someone who hasn't seen it, one playthrough, a description of what happened and why — plus a separate pass on the captions for the original, checked against the audio they were written to carry.

Frequently asked questions

Why is muting a spot not enough to design a sound-off version?

Muting tells you what breaks, not what to build in its place. It can show that information is missing or that an action now reads differently, but the rebuild usually needs a visible cause, a re-blocked action, or a line of text placed with room to be read. Labeling every sound instead tends to produce a transcript laid over an edit that was cut for the ear.

What three jobs can a soundtrack be doing that a silent version must account for?

Speech can supply a fact the story turns on. Sound can identify a cause you cannot see, such as a chime, knock, or door, because a reaction alone says something happened but not what it was. Music can change how a retained action reads; without it the same shot can mean something else, not just less.

In the invented 20-second office scene, what goes wrong in the muted cut and what is the smallest fix?

The phone screen lights at 0:04 and the worker hides the box by 0:06, but the audience cannot read the message. The door opens at 0:14, so the reason for hiding lands late and the hiding has been filed under the wrong genre, such as trouble or evidence. The smallest fix is to make the phone legible at the top of the scene: same object, same beat, cause already in frame. That must survive every delivered frame shape and not steal the beat from the worker's face. If it fails, move the cause to the door and let the silhouette carry it, accepting that 'someone is outside' replaces 'your visitor has arrived.'

What is the difference in cost between an in-shot legible screen and a text overlay?

A message the audience reads inside the shot is a photographed object; making it legible is a change of framing, not a new layer. A line or graphic laid over the picture is a layer that spends a reading clock the original cut never paid for. Reading and watching need separate moments: move the text earlier so reading finishes before the action, or cut the information down until it clears the beat. Do not shrink the type, because that does not create reading time. Check the actual delivery, frame shape, duration, and screen size.

How should the silent redesign be reviewed, and how do captions fit into the work?

Get someone who has not seen it, mute the version, play it once, and ask them to describe what happened and why, not whether it felt clear. Listen for which facts they volunteer and where they place them in sequence; if they supply an unintended cause, that is the under-specified slot filling itself in. Captions are a different deliverable: W3C's Web Accessibility Initiative guidance, which carries a September 2024 update, describes captions as synchronized text for speech and meaningful non-speech audio and distinguishes captions from subtitles. That definition does not validate a creative redesign or certify accessibility. Captions assume the audio exists and describe it; they do not fix an unseen cause, and they must be turned on. Keep the original mix for captioning, and treat the sound-off version as an additional deliverable, not a replacement.

Further reading

• Voiceover or On-Screen Text: Which Should Carry the Message? — Read this when deciding whether voice or text carries the message, and when a required silent version must stand as its own edit.
• Add Captions to a Film Pitch Clip and Deliver the Right Version — Read this when the captions for the original need their own pass, timed against the cut and delivered in a route the recipient can open.
• Make a Descriptive Transcript of a Commercial Pitch Sample — Read this when a reader needs the visual facts and sounds in reveal order, not just synchronized text for audio that already exists.