What it asks for, in plain English
Any recorded video that has sound needs captions: text shown in time with the video that gives the speech and the sounds that matter, such as a door slamming, laughter or a phone ringing, and who is speaking when that is not obvious.
Captions are not the same as translated subtitles. Subtitles assume you can hear the sound effects and the tone; captions do not.
Automatically generated captions count only after someone has checked and corrected them. Raw auto-captions routinely get names, numbers and technical terms wrong. The one exception is a video that is a labelled alternative to text already on the page.
This is an independent explanation. The binding wording is the W3C Recommendation, and the W3C's own Understanding 1.2.2 page has the full intent, techniques and definitions.
Who it affects
Someone who is deaf or hard of hearing sees people talking and has no way to know what was said or decided.
A parent watching a tutorial on mute next to a sleeping child follows the steps from the captions and never needs to turn the sound on.
Someone whose first language is not English reads along to catch words the speaker's accent or pace makes hard to pick out.
Example
<video controls src="onboarding.mp4" width="640"></video>
The video plays, but there is no captions track and nothing in the player to turn one on.
<video controls src="onboarding.mp4" width="640">
<track kind="captions" src="onboarding.en.vtt" srclang="en" label="English" default>
</video>
A WebVTT captions file is attached and the player shows a CC button. The file still needs to be accurate.
How to check it
- Play each video with the sound off and captions on. You should be able to follow what is said and what is happening.
- Spot-check a minute in the middle against the audio: names, numbers and product terms are where auto-captions go wrong.
- Check that meaningful sounds and speaker changes are captioned, not just the dialogue.
- For embedded players (YouTube, Vimeo), confirm captions are uploaded for that video rather than relying on the platform's automatic ones.
What Websight checks for 1.2.2
Websight runs this automated check for 1.2.2:
Websight's video caption check flags a <video> element that has no captions track. It cannot judge whether the captions are accurate, and it cannot see captions burned into the picture or provided inside an embedded third-party player, so those still need a person. This is on the manual checklist.
Practical tips
- Start from the auto-captions, then correct them; it is far quicker than typing from scratch.
- Include speaker names in square brackets when two or more people talk.
- Keep each caption to two lines and give readers time; if the speech is fast, that is a reason to edit the video, not to cram the captions.
Other criteria in guideline 1.2, Time-based Media
- 1.2.1 Audio-only and Video-only (Prerecorded) (level A)
- 1.2.3 Audio Description or Media Alternative (Prerecorded) (level A)
- 1.2.4 Captions (Live) (level AA)
- 1.2.5 Audio Description (Prerecorded) (level AA)