Synchronized Audio Description

Synchronized captions are a text version of the dialogue and sounds of a video, displayed in real time in the player. They are not limited to dialogue: they identify who is speaking and describe significant sounds such as music or a door slam. WCAG requires them at Level A for pre-recorded videos (criterion 1.2.2).

Your video has captions. But do they describe the door slam, the tense music, the character's sigh? If the answer is no, a deaf person is missing part of the message.

#Captions or subtitles: the terminology

In English, two distinct words exist. Subtitles translate dialogue for an audience that can hear. Captions transcribe all audio (speech, speaker identification, significant sounds) for an audience that cannot hear.

French has only one word for both. This ambiguity is reflected even in web projects: many teams deliver subtitles thinking they have provided captions.

In HTML, the <track> tag makes the distinction via the kind attribute:

<!-- Accessibility captions -->
<track kind="captions" src="captions-en.vtt" srclang="en" label="English (Captions)" default>
 
<!-- Translation subtitles -->
<track kind="subtitles" src="subtitles-fr.vtt" srclang="fr" label="Français">

kind="captions" signals to the browser and assistive technologies that the track contains the full audio information. WCAG criterion 1.2.2 requires captions, not simple subtitles.

#The automatic captioning trap

Many teams enable automatic captioning on their video platform and move on. Two problems.

Automatic captioning does not distinguish speakers. In a two-voice exchange, it's impossible to know who is speaking. It also ignores non-verbal sounds. The WebVTT format allows you to describe them:

WEBVTT
 
00:00:12.000 --> 00:00:15.500
[Tense music]
 
00:00:16.000 --> 00:00:18.200
<v Marie>We can't keep going like this.
 
00:00:19.000 --> 00:00:20.500
<v Paul>I know.

The <v> tag identifies the speaker. Brackets describe sounds. Without these indicators, your captions are disguised translation subtitles.

Second problem: accuracy. According to NoteLM.ai tests, YouTube's automatic captions achieve 85 to 95% accuracy under the best conditions. Proper nouns and technical vocabulary are the first to suffer, especially with regional accents. Human review remains essential.

#In summary

One person in eight presents a functional hearing limitation, according to U.S. government data cited by WebAIM. Synchronized captions go beyond dialogue text: they identify speakers and describe significant sounds. In HTML, use kind="captions" (not kind="subtitles"), provide a human-reviewed WebVTT file, and verify that every sound necessary for comprehension is included.

Back to glossary

Share this article

Learn more

Synchronized Audio Description

Synchronized audio description is a voice narration inserted during pauses in a video to describe visual elements that the soundtrack does not convey. It enables people who are blind or visually impaired to follow the action, scene changes, and on-screen text. WCAG requires it at level AA for all prerecorded video content.

Time-based media

Time-based media is content whose understanding depends on time flowing: audio file, video, or both combined. In accessibility, each type of time-based media requires specific alternatives: transcription, captions, audio description. WCAG devotes nine success criteria to it (guideline 1.2).

Transcript

A transcript is the written version of audio or video content, accessible independently from the player. It reproduces the spoken words, identifies speakers, describes significant sounds, and for videos, the visual information necessary for understanding. WCAG requires it from level A for all prerecorded audio-only content (criterion 1.2.1).