A video starts by itself in a feed, most often with the sound off. LinkedIn writes it in its help pages: some videos play automatically with the sound off as people scroll down the page. During those first seconds, what you say exists only if it can be read.
People watch with the sound off, and say so
Solid figures on the subject are rare, and it has to be said where they come from.
The most quoted comes from a study by Verizon Media and Publicis Media, run in April 2019. Among the people surveyed, 69% say they watch videos with the sound off in public places, and 25% in private places. Above all, 80% say they are more likely to watch a video to the end if it has captions. Two reservations: these are self-reported habits, and the study comes from a seller of advertising space and a media agency, which published neither their sample nor their method.
Other sources point the same way, on other screens.
- Ofcom, the British broadcasting regulator, noted in 2023 that 42% of adults in the United Kingdom use subtitles when they watch television, streaming included. Among 16-to-24-year-olds, the figure is 63%, against 36% among those aged 55 and over.
- YouGov surveyed 1,000 American adults in 2023: 38% prefer to watch television with subtitles on, even in a language they know. Among the under-30s, the figure is 63%.
These two figures are about television and streaming, not about social networks. They say something useful all the same: reading what you hear has become a habit, and it is stronger among the youngest.
Who captions are for
First, those who cannot hear, or hear poorly. More than 1.5 billion people live with hearing loss, which is nearly 20% of the world’s population, according to the World Health Organization. For them, a video without captions is a video without content.
That is why the web’s accessibility rules make captions a basic requirement. Criterion 1.2.2 of WCAG 2.2, at level A, the first level, asks for captions for all prerecorded audio content in a video.
Then, everyone else. According to the Verizon Media study, 80% of the people who use captions are neither deaf nor hard of hearing. They are in a quiet place, without earphones, or busy doing something else.
What captions do not replace
It would be wrong to conclude that sound no longer matters. A study published in PLOS ONE in 2024 tracked the gaze of 168 people watching captioned videos, with and without the sound. Without the sound, they understood less well, made more effort and took less pleasure in watching.
So captions have two distinct jobs:
- holding the person who scrolls by with the sound off, long enough for them to decide to stay;
- helping the person who is listening to follow, above all when the subject is technical or when you talk fast.
They do not excuse you from speaking well or from having clean sound. A short that works is one that works with the sound on, and can be understood without it.
Automatic or burned in
There are two kinds of captions, and they do not do the same job.
| Automatic captions | Burned-in captions | |
|---|---|---|
| Where they come from | generated by the platform | made in the edit, before uploading |
| Where they are | laid over the video | in the picture itself |
| The viewer can hide them | yes | no |
| Their look | the platform’s | yours |
| From one platform to another | to be redone every time | identical everywhere |
Automatic captions exist on the four big platforms. LinkedIn offers them in ten languages, French and English among them. YouTube creates them for Shorts as it does for long videos. Instagram turns them on through a setting for each reel, from the app. TikTok generates them for uploaded videos.
They are a help, on two conditions. You have to read them through: YouTube itself recommends checking what was wrongly transcribed, and a mangled name makes a poor impression. And you have to accept not choosing: neither the font, nor the size, nor the place where they are displayed.
Burned-in captions are part of the video. Nobody can hide them, they are displayed even where automatic captions are not offered, and they carry your style. LinkedIn, for that matter, advises advertisers to burn in their captions.
For a short, they are the ones that count: they are the first thing people read.
Seven rules for captions people read
1. Few words at a time
A group of words, one or two short lines. The viewer has to read at a glance, without their eyes travelling across the screen.
2. Break where the voice breaks
A caption stops where the sentence takes a breath. “Nobody reads / a deck twice” reads better than “Nobody reads a / deck twice”.
3. In time with the voice
The word appears when it is spoken. A lag, even a slight one, bothers anyone who is listening while reading.
4. Big and high in contrast
The text is read on a phone, sometimes in broad daylight. Thick, light letters, with an outline or a dark background: a video changes in brightness, and the text has to stay readable on every one of its frames.
5. Out of the covered zones
The platforms display their interface at the bottom of the picture and on its right. Captions placed there disappear under the description. Put them in the centre, a little above or a little below the face. The exact margins are in The 9:16 vertical video format.
6. Faithful, but cleaned up
Write what is said, without the “ums”, the repetitions or the false starts. If the edit took them out of the sound, they have no business in the text.
7. Read through
Names, acronyms, trade terms and figures are what automatic transcription gets wrong most often, and what your audience notices first.
The line above the video
Captions say what you are saying. They do not say what you are talking about, at least not before the end of the first sentence.
A fixed line of text, at the top of the picture, solves that problem: it states the question the video answers. “How do you pitch in one call?” The viewer knows within a second whether the subject concerns them, with or without the sound.
It is also the simplest way of making a passage stand on its own, when the question was asked by someone who is not heard in the clip. We talk about it in How to choose video clips.
What Dailies does
Dailies records the calls you hold in Chrome, in French or in English, and cuts vertical shorts from them. Its extension’s interface is in French. Each short arrives with its captions burned in, in one of the fifteen styles on offer: you pick one, and all your shorts follow it. The silences, the hesitations and the false starts are taken out of the picture, the sound and the text. A line above each short states the question it answers.
Key takeaways
- A video often starts with the sound off: what cannot be read does not exist.
- Captions serve first the people who hear poorly, and everyone who watches without listening.
- They do not replace sound: without it, people understand less well.
- Burned-in captions are the only ones whose look you control, and they follow you everywhere.
- Few words at a time, big, high in contrast, in time with the voice, out of the covered zones, and read through.
Sources
- Verizon Media and Publicis Media, Relationship of Video, Sound and Captions Insights Study, April 2019: self-reported habits, sample and method not published.
- Ofcom, Review of Public Service Media (2019–23), 2024: adults in the United Kingdom, 2023 data.
- YouGov, survey on subtitles on television, 2023: 1,000 adults in the United States.
- Szarkowska and others, study published in PLOS ONE, 2024: 168 participants.
- World Health Organization, Hearing loss.
- W3C, WCAG 2.2, criterion 1.2.2, 2024.
- Help pages of LinkedIn, YouTube, Instagram and TikTok on automatic captions, read on 1 October 2026; LinkedIn’s pages on autoplay and its tips for advertisers.