Caption live video streams in real time
Live video needs captions while it is live, and G9 is the transcription half of that job. A stenographer, a re-voicing captioner, or a trained speech-to-text setup turns speech into text as it happens. Read the status carefully, because it comes with a condition. W3C counts G9 as sufficient only when it is paired with a delivery technique, either open captions burned into the picture or closed captions the viewer switches on. G9 produces the words and something else has to put them on the screen. One more constraint settles a question that comes up often. The captions have to be in the same language as the speech, so a machine translation into another language is not captioning under this rule.
How we find it in an audit
We watch a live stream, because something that exists only while it is happening cannot be tested any other way. A reviewer joins, checks the captions appear within a couple of seconds of the speech, and reads them against what is actually being said, with names, figures and jargon getting the most attention. The delivery route goes into the record too, open or closed, since the technique is only complete with one of them. Where a client runs live events, we ask what the booking arrangement is, because live captioning is a contract rather than a code change.
How affected users experience it
A deaf employee joining the quarterly all-hands without captions is in the room and not in the meeting. They watch faces react to something and reconstruct the topic from the chat afterwards. Automatic captions with nobody correcting them cause a different problem, because a name or a figure that comes through wrong reads as confident and true. And a caption arriving twenty seconds late has stopped being a caption. It is a transcript turning up during the next answer.
Passes vs. fails
Passes
A real-time captioner joins every stream, and the words appear in the player within a couple of seconds of being spoken.
Fails
A quarterly all-hands streams with no captions, so deaf employees follow along by guesswork and by reading the chat.
Other ways to satisfy this rule
5 guides on this site are filed under 1.2.4 Captions (Live). W3C lists this one as sufficient for that rule only alongside G93, so the pair is what passes and neither half does on its own.
This guide is our interpretation of W3C technique G9: Creating captions for live synchronized media. W3C publishes its techniques as guidance rather than as the standard, and says so on every one of them. The success criterion is what conformance is measured against, and a technique is one documented way to meet it.