Audio-only and Video-only (Prerecorded)
A recording with only sound needs a text alternative carrying the same information. A recording with only pictures needs either that same text alternative or an audio track telling you what is on screen. That is the whole of WCAG 1.2.1, and it has one exit. If the recording adds nothing that is not already written on the page, and you label it as an alternative to that text, the rule does not apply at all. The text version is more than a rough transcript. It runs in the recording's own order, names who is speaking, and notes the sounds that carry meaning. Where the recording asks you to do something, it gives you a working way to do it.
Why it matters
Single-channel media puts everything it has to say into one sense. When that sense is unavailable, the content is simply not there. A transcript hands a deaf reader the whole episode in a form they can take at their own pace. For a silent video there is a choice worth making on purpose. Written steps underneath the player work. Narrating the video works better, because then a blind viewer and a sighted viewer are watching the same thing at the same moment instead of reading a separate document. Narration also helps people who find a moving picture hard to follow. And a transcript makes the recording searchable, quotable, and reachable by anyone who would rather read than listen.
Who this rule protects
Deaf and hard-of-hearing users cannot follow a podcast without a transcript. Blind users cannot follow a silent demo without narration or written steps. Deafblind users depend on the text version, because braille is the only route into either one. Anyone in a quiet office or a loud train reads the transcript too.
How to check it yourself
- Find every audio-only player on the site and confirm a full text alternative sits on the page or one click away.
- Read the text version against the recording. It has to carry the dialogue and every sound that matters, either word for word or in essence. What fails is a summary that drops information, not one that tightens the wording.
- Check the text version names who is speaking whenever more than one person talks. It should also note the applause, the laughter and the questions from the room.
- Play every silent video and confirm a written description or a spoken track covers everything that happens on screen.
- Where a recording tells the listener to do something, check the text version gives them a working route to the same outcome, usually a link.
- Ask whether the media is really an alternative to text already on the page. If it is, and you have labeled it as one, it needs nothing further.
Failures we see most often
- An episode page publishes show notes and calls them a transcript, so most of the conversation exists nowhere in text.
- A silent product demo or an animated GIF ships with no written steps and no narration.
- A transcript runs the dialogue together with no speaker names, so a two-person interview reads as one long monologue.
- A transcript drops the sounds that carried the story, such as the laughter telling you a line was a joke.
- A video version of a help article is never labeled as an alternative to the text, so the exception does not apply and the video owes a text alternative or a narrated track of its own.
- The text alternative is a file name or a placeholder, which is not an alternative to anything.
- A long description exists and carries different information from the recording, which leaves the recording still uncovered.
Who this one is for
Read from this rule's own note above, so the grouping and the note cannot disagree.
- Blind and screen reader userspeople who cannot see the screen
- Deaf and hard of hearingpeople who cannot hear the audio
How this one is tested
We list 2 ACT rules against 1.2.1. 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.
- Audio element content has text alternativeA tool finds candidates, you decide
- Video element visual-only content has accessible alternativeA tool finds candidates, you decide
How to fix it
- Publish a full text alternative next to every audio-only recording. Speech-to-text software plus a human pass to fix names, terms and speaker labels is usually enough.
- Write the text version in the recording's own order. Speaker names, the sounds that matter, and a link wherever the recording tells somebody to go somewhere.
- For a silent clip, record narration describing what happens, and say plainly that the track is the description. That description does not then need captioning of its own, because it is describing what is already on screen.
- Where narration is not possible, publish numbered written steps beside the video and keep them in the order the video uses.
- If the recording only repeats text already on the page, label it as an alternative to that text. That is the whole job.
Step-by-step fix guides (6)
Passes vs. fails
Passes
The same video ships with numbered written steps underneath it, and the narrator names each measurement out loud, so every shopper can follow the method.
Fails
A "How to measure your ring size" video plays with no sound and no transcript, so a blind shopper gets nothing from it.
In audits and lawsuits
In our audits this shows up most often in two shapes. Episode pages with show notes and no transcript, and silent how-to clips with nothing written to explain them. A scanner can tell you a transcript exists. It cannot tell you whether the transcript matches the recording, which is why every test rule written for this criterion needs a person. One timing point catches people out. A recording of a live event becomes prerecorded the moment you publish the file, so the clock on this rule starts at publication rather than at the event. A live feed is a different question. This rule is prerecorded only, so a silent webcam or traffic camera running live needs nothing beyond a label saying what it is, which belongs to 1.1.1 instead.