Most People Watch Your Video With the Sound Turned Off

Put her words on the screen. A testimonial nobody can hear is a testimonial nobody receives.

Think about where people actually scroll: on a bus. In a waiting room. In an open-plan office. In bed, next to someone asleep. Sound off, every time.

Your thirty-second testimonial is a person talking. Take the sound away and it is a stranger silently moving her mouth at a phone. Everything you collected is gone.

Subtitles are not a nice extra for accessibility. For this particular kind of video, they are the entire delivery mechanism.

The number I am not going to give you

You will have seen “85% of video is watched without sound”. Or 80%. Or 92%, depending on which marketing blog you land on.

Chase any of them and you find a blog citing a blog citing nothing. So I will not repeat one — a number without a source is a decoration, not a fact, and you should not put one in your own marketing either.

What you do not need a statistic for: look at where you are, right now, and whether your own sound is on. Then think about the last time you scrolled on a bus. The behaviour is not in dispute; only the decimal place is.

Picture a physio clinic, thirty seconds, no sound

Say a physiotherapist films a patient in the corner of the treatment room. The man has come back after a knee that kept him off five-a-side for a year, and he is genuinely pleased. He talks for thirty seconds — a bit rambling, grateful, the way people are when something has actually worked.

Now watch that clip the way it will really be seen: muted, thumbed past on a sofa at nine at night. Without text, it is a middle-aged man in a polo shirt moving his mouth in a room full of equipment. Nothing about it says knee, says year off the pitch, says thank you. The one asset — a real person saying a real thing — never arrives.

Put his words on the screen and the same thirty seconds does its whole job in silence. The viewer reads “couldn’t play football for a year”, reads “back on the pitch in March”, and decides to turn the sound on to hear him say it. The subtitle bought you the second chance. A patient talking about a result they can feel is one of the most persuasive things a clinic owns — but only if it can be received with the sound off.

Subtitles are where the tidying sneaks in

Now the important part, and it is the reason this article exists.

The subtitles are text. Text feels editable in a way that audio does not. Nobody would dream of splicing a customer’s sentence in the audio — but almost everybody, given an auto-generated caption track full of “ehm” and half-finished sentences, quietly cleans it up.

You fix the grammar. You delete the “ehm”. You join the sentence she abandoned to the one she started instead. It takes two minutes and it looks so much better.

And you have just rewritten what she said.

Her words go out exactly as she spoke them — and that includes the subtitles. Not tidied, not shortened, not improved. For a viewer with the sound off, the subtitle is the testimonial: it is the only version of her they will ever receive. A cleaned-up subtitle track is not a cosmetic improvement to a real testimonial. To most of your audience, it is a fabricated one.

This is the trap that catches people who have understood everything else. The caption is yours; her words are hers — and the subtitle is her words, not your caption.

Fix the errors, do not fix the person

One legitimate exception, and hold the line exactly here.

Automatic transcription gets things wrong. It hears “balayage” as “ballet age”. It renders a name as gibberish. It mangles a word it has never encountered.

Correcting a transcription error is restoring what she said. Correcting her is replacing what she said. The test is simple: would she recognise this as her sentence, or as a better one?

• She said “balayage”, the machine typed “ballet age” → fix it. That is a repair.
• She said “it’s, ehm, it’s honestly so much better than — yeah, than I expected” and you typed “It’s honestly so much better than I expected” → do not. That is a rewrite, and it is the one that costs you.

Sound still matters, because half of them will turn it on

Do not read all this as “sound does not count”.

The people who are interested — the ones actually considering booking — often do turn the sound on, and that is when they hear the thing that convinces them: the warmth, the hesitation, the small laugh, the real intonation of a person who is pleased and slightly self-conscious.

The subtitle gets her past the scroll. The voice closes it.

Which is why sound quality is still the first thing to get right at the moment of recording: get within an arm’s length, kill the hairdryer, point the bottom of the phone at her. Sound matters more than picture, and no subtitle rescues a recording nobody can bear to listen to.

Burned-in, or the platform’s own captions?

Instagram and TikTok can generate captions for you, on their side. It is quicker, and it saves you a step. But two things are worth knowing before you lean on it.

First, the viewer can often switch platform captions off, and on some feeds they do not come on by default. Captions burned into the video itself — baked into the pixels — are always there, on every platform, for every viewer. Nothing to toggle, nothing to depend on.

Second, and this is the one that matters here: the platform’s auto-captions are auto-generated, which means they carry exactly the same errors — and exactly the same temptation to tidy — as any other transcript. “Balayage” comes out “ballet age” whether the machine typing it is yours or Instagram’s. So the rule does not change with the tool. Read what it produced. Fix the errors. Leave the person alone.

And if you post the same clip to several platforms, burned-in captions travel with it and look the same everywhere. Platform captions have to be redone, and re-checked, on each one — which is three more chances to quietly “improve” a sentence she never said.

Captions also keep people watching

There is a quieter reason to caption, beyond the person on the bus. A video that makes sense with the sound off is a video people stay with for longer, because they do not scroll away in the first second wondering what it is. And staying is the single thing every platform reads as “show this to more people”.

So you are not adding captions to please an algorithm. You are adding them so a human stops and reads — and the algorithm is only ever counting how many humans stopped.

Keep them readable

Practical, and quickly:

• Big enough to read on a phone held at arm’s length. If you have to squint on your own screen, nobody else is reading it.
• High contrast. White text with a dark outline survives any background.
• Out of the danger zones. The bottom of a Reel or a TikTok is covered by the caption, the username and the buttons. Text placed there is invisible on the actual platform, even though it looks fine in the editor.
• No animated word-by-word karaoke on a testimonial. It is the visual language of a produced advert, and it fights the one thing your video has going for it, which is that nobody produced it.

Turn your own sound off and watch it

Before you publish, do this once: mute the video and watch it as a stranger would.

Can you follow what she is saying? Is the text readable, on screen long enough, not hidden behind a button?

If yes, you have a testimonial. If no, you have thirty seconds of a woman mouthing something at a phone, and everything you did to get it was wasted at the last step.

And whatever the subtitles say, make sure it is what she said — hesitations and all. That is the only version worth putting on screen.