Audio-only (Live)
Live audio with no picture needs a text alternative carrying the same information, available while the audio is running. A live captioning service is the usual answer. There is a much cheaper one W3C also documents. If the broadcast follows a prepared script and does not wander off it, publishing that script is enough.
Why it matters
Live audio with no text is closed to deaf and hard of hearing listeners, and unlike a recording there is no catching up later, because the event happens once. Note that this covers broadcast rather than conversation. W3C says plainly that it was never meant to require captioning on two-way calls between people, and the definition of audio-only rules out anything interactive anyway.
Who this rule protects
Deaf and hard of hearing listeners need this most, along with anybody listening somewhere they cannot turn the sound on.
How to check it yourself
- List any live audio-only streams you publish, and leave two-way calls out, because the criterion does not reach them.
- Ask whether the audio follows a set script. If it does, and the speaker sticks to it, publishing the script is a documented route.
- Where there is no script, check that a text alternative runs while the audio does, not afterward.
- Sample the output on names, numbers and technical terms, because a transcript that diverges markedly from the event does not count.
Failures we see most often
- A transcript gets published after the event and called a live alternative.
- Automatic captioning runs unattended on content full of proper nouns it keeps getting wrong.
- A scripted broadcast wanders off the script, so the published script no longer matches what listeners heard.
Who this one is for
Read from this rule's own note above, so the grouping and the note cannot disagree.
- Deaf and hard of hearingpeople who cannot hear the audio
How to fix it
- Book a live captioning service for any scheduled unscripted event.
- Where the event is scripted, link the script from the page carrying the audio, and hold the speaker to it.
- Where automation is the only option, staff it with somebody correcting the feed in real time.
- Publish the corrected transcript afterward too. That is what criterion 1.2.1 asks of the recording once it stops being live, and 1.2.1 is Level A.
Step-by-step fix guides (3)
Passes vs. fails
Passes
A company streams a prepared statement and publishes the written statement at the same moment, which is the route W3C's own example takes.
Fails
A live audio AMA posts its transcript three days after the event ends.
In audits and lawsuits
This is Level AAA and the media criterion with the widest gap between its cheapest route and its most expensive one. A scripted broadcast can meet it for the price of publishing the script. An unscripted one needs a live captioner, which W3C describes precisely. A trained operator on a stenographic keyboard, typing with a small delay, adding notes on any non-speech sound the listener needs. What W3C rules out is doing it badly. Untrained operators, or a transcript that differs markedly from what actually happens, do not meet the criterion.