A large share of video advertising in social feeds is watched without sound. That single fact determines how the format is written, shot and edited.
Autoplay starts muted by default
Platforms mute autoplaying video because unexpected sound in a public place drives people away from the app.
Unmuting requires a deliberate action, and most viewers do not take it for content they have not yet decided to watch.
The ad therefore begins in a state where the entire audio track is unavailable, and it has to earn the tap before sound becomes relevant.
Anything spoken is effectively unsaid
A voiceover that carries the proposition delivers it to the minority who unmuted. For everyone else the ad ran without its argument.
This is why silent ads so often fail for reasons the creative team cannot see, since every internal review is conducted with sound on.
The correction is to review muted, which reliably changes which cuts are approved.
It also changes the brief, because a proposition that only exists as a spoken line has to be rebuilt as something the picture can carry on its own.
Captions are a delivery mechanism, not an accessibility extra
Burned-in text carries the spoken content to viewers who never hear it, which is most of them.
Platform-generated captions are inconsistent and can be switched off, so text that matters belongs in the video file itself.
Placement matters as much as presence, because interface elements overlay the edges of the frame differently on each platform and can obscure text placed there.
Sound still does work when it arrives
Sound off is the default, not the whole audience. Viewers who unmute are the most engaged and the most likely to convert.
The audio track should therefore reward that decision rather than merely repeating what the captions already said.
Music and sound design also carry brand association across formats where no visual system exists, which makes consistency there more valuable than it appears.
Where the assumption reverses
Connected television is watched with sound on, at a distance, usually by more than one person. Every constraint above inverts.
Text sized for a phone is unreadable across a room, and the rapid cutting that holds attention in a feed reads as frantic on a large screen.
Running one master edit across both environments therefore compromises in both directions, which is the most common avoidable defect in cross-channel video plans.