Watching Video Players: Captions That Lag the Audio
Captions that trail the spoken words by two seconds are worse than no captions for a deaf viewer. Product videos are the worst offenders on storefronts.
A product video plays on a store page. The narrator says the jacket is waterproof, and two seconds later the caption catches up. A deaf viewer reads the caption after the shot has changed, and by the third mismatch she has stopped trusting the captions entirely. Lagging captions are not a minor sync issue. They are the difference between a video a deaf customer can use and one she cannot.
Product videos are the worst offenders because they are produced fast, captioned by auto-generation, and never checked against the audio. The fix is a timing check, and it takes less time than the video shoot did.
Why captions drift
Auto-generated captions are usually accurate on words and wrong on timing. The speech recognition engine timestamps each word against the audio, and the player renders captions in fixed blocks, so a fast speaker's words bunch up and a slow speaker's words lag. Edited videos make it worse: a cut that removes two seconds of audio without re-timing the caption track leaves every caption after the cut two seconds late.
WCAG 2.2 criterion 1.2.2 requires captions for prerecorded audio in synchronised media at Level A. Synchronised is doing real work in that sentence. A caption track that trails the audio by a visible margin is not synchronised, and a court or an auditor will not accept it as meeting the criterion.
How to check it in five minutes
Play the video with the sound off and the captions on, and watch whether each caption matches the shot and the speaker's mouth. Then play it with sound on and read along. If you find yourself reading a sentence after the speaker has moved on, the track is lagging. Check the first ten seconds and the last ten seconds, because drift accumulates and the end of a video is usually the worst part.
Check the player controls too. The captions toggle must be reachable by keyboard, and turning captions on must not require a mouse. A player whose caption button is a div with no keyboard support fails two criteria at once, and the caption fix is pointless if nobody can turn captions on.
Fixing the track
Re-time the captions in the caption editor rather than re-recording the video. Most caption formats let you drag a caption block to the right timestamp, and a two second drift is usually one adjustment, not a rebuild. If the auto-generated track is the problem, regenerate it from the final edit, not from the rough cut, and proofread the result for the words that matter: sizes, prices, and product names.
For videos that change often, such as seasonal product demos, put the caption check in the publishing checklist. The person who uploads the video watches it once with captions on before it goes live. That single viewing catches the drift that a scan never will.
Related reading: Watching Live Chat Widgets for Keyboard Traps and Focus Loss, Watching Third Party Payment Buttons: Iframes, Titles, and Tab Order, Watching Autoplay Video Backgrounds and Their Pause Controls.
Frequently asked questions
Do captions have to be perfectly synchronised?
WCAG 1.2.2 requires captions for prerecorded audio in synchronised media. A caption track that trails the audio by a visible margin is not synchronised, and the usual test is whether a viewer reading the captions can follow the video without the sound on.
Why do auto-generated captions drift out of sync?
Speech recognition timestamps words against the audio, and players render captions in fixed blocks, so fast speech bunches up and slow speech lags. Edits that cut audio without re-timing the caption track leave every caption after the cut late by the same amount.
What is the quickest caption check for a product video?
Watch it once with the sound off and captions on, and once with sound on while reading along. Check the first and last ten seconds, where drift is worst, and confirm the captions toggle itself is reachable by keyboard.
Does the caption button need to be keyboard accessible?
Yes. A caption toggle that only works with a mouse fails keyboard access, and the captions are useless to the people who need them most if they cannot be turned on. The toggle must be a real button in the tab order.