Skip to main content
WCAGrules
Quick navigation

Captions (Prerecorded)

Every prerecorded video with sound needs synchronized captions. Captions are not a transcript link, and they are not subtitles either. Subtitles carry the dialogue. Captions carry the dialogue plus everything else you would need to hear to follow the video, meaning sound effects, music, laughter, who is speaking and where they are. They can be a caption track a viewer switches on. They can also be burned into the picture so nobody can switch them off. Both of those pass. The one exit is the one running through this whole guideline. If the video adds nothing that is not already written on the page, and you label it as an alternative to that text, no captions are required.

Why it matters

Captions carry the whole soundtrack for deaf and hard-of-hearing viewers. That soundtrack exists for them in no other form. There is a second group named in the standard itself, which is people who take written information in better than spoken. Plenty of hearing viewers turn captions on too, watching muted in an open office or beside a sleeping child. One thing captions are not is a translation. A subtitle track in another language still leaves out the sounds, so it is a service to your international audience rather than an accommodation for a deaf one.

Who this rule protects

Deaf and hard-of-hearing viewers rely on captions to follow the speech and everything else in the mix. People who take written information in better than spoken rely on the same text. So does anyone watching with the sound off, which on a phone in public is most people, most of the time.

How to check it yourself

  1. Play every video that has sound and confirm captions exist and run in step with the speech. Turn them on where the player has a switch. Where they are burned into the picture and cannot be turned off, that passes just as well.
  2. Check the captions carry all the dialogue, either word for word or in essence, plus every sound that matters. Condensing a long sentence is fine. Losing one is not.
  3. Check speakers are named whenever more than one person talks. Meaningful sound effects should appear in brackets, written as [door slams] or [phone ringing].
  4. Spot-check auto-generated captions against the audio in two or three minute stretches. Listen hardest for brand names, prices, numbers and the word not.
  5. For music with no words, check the captions identify it well enough to be worth reading. Title, movement and composer, with the melody described, is the pattern the standard illustrates.
  6. Where a video carries nothing the page does not already say, check it is labeled as an alternative to that text. Without the label, the exception does not apply and captions are due.

Failures we see most often

  • A marketing video is uploaded with no caption track at all.
  • Auto-captions are published unreviewed, so brand names, prices and negations come out wrong on screen.
  • Captions carry the dialogue and drop the sound effects, so a viewer who cannot hear the alarm never learns there was one.
  • No speaker label ever appears, so a panel discussion reads as one person contradicting themselves.
  • A video is presented as the alternative to the page's own text and never labeled as one, so the exception it relies on does not apply.
  • A video claims that exception while showing more than the page says, which means captions were due all along.
  • Burned-in captions sit over the part of the picture the viewer needs to see, which is the one presentation rule the captions definition sets.

Who this one is for

Read from this rule's own note above, so the grouping and the note cannot disagree.

How this one is tested

We list 1 ACT rule against 1.2.2. Each one defines exactly what a checker looks at, which is how automated tools decide what to flag. Each one also checks a slice, so passing every rule here is not the same as meeting the criterion, and a rule can be proposed rather than approved or need a person to finish it. The note beside each says which.

How to fix it

  • Add a caption track to every video with speech. In HTML that is a track element pointing at a WebVTT file, which is the format browsers actually read, so convert an SRT export before you ship it. Correcting the machine output before you publish costs far less than correcting it afterwards.
  • Burn the captions into the picture where a track cannot travel with the file, which is the usual case for social video. Always-on captions pass this rule as cleanly as switchable ones.
  • Caption the sounds, not only the words. Bracket the effects, name the speakers, and mark music with its title and its mood.
  • Keep captions clear of anything on screen the viewer needs to read. Whether the captions themselves are legible against the footage is a contrast question rather than a captioning one, and it is measured the way body text is.
  • Two lines at a time, timed within a beat of the speech, is captioning industry convention rather than a WCAG requirement. Follow it anyway. It is what makes captions readable at speaking pace.
Step-by-step fix guides (8)

Passes vs. fails

A general illustration of the pattern rather than a test of 1.2.2. Passes: reviewed captions, in sync. Fails: speech with no captions.

Passes

The same video ships with reviewed captions that name both speakers and mark the applause, so anyone can watch it silently, anywhere.

Fails

A founder-story video ships with no captions, so muted viewers scroll past it and deaf viewers cannot watch it at all.

In audits and lawsuits

In our audits the finding is rarely zero captions. It is auto-generated captions nobody read afterwards. The standard is not kind about those, because a caption track only counts once it is confirmed accurate. Its worked example is a dropped negation, where a spoken "you should not preheat the oven" arrived on screen as "you should know to preheat the oven". Deliberate condensation is allowed and expected, since a viewer reading at speaking pace needs the shorter line. Dropping a sentence or garbling a brand name is not. The standard also acknowledges the daily-video problem and still fails you for it. A newsroom is left choosing between holding the story and publishing something deaf viewers cannot watch.

Go somewhere useful

Find tools, resources and your workspace.

29 destinations