WCAG 1.2.5 — Audio Description (Prerecorded)

What this requires

For prerecorded synchronised media at Level AA, an audio description track is required. The description narrates the on-screen visuals during the natural pauses in dialogue: scene changes, on-screen text, visual gestures, character actions — anything a blind user would otherwise miss. Unlike 1.2.3, a text transcript is not a substitute at this level.

How AI coding tools fail this

When asked to embed a hero video or a product walkthrough, AI tools generate a single audio track and stop. The audio-description track — which usually exists as a separate file or a separate selectable track on the platform — never gets wired in. Even when the project has a described version produced, the embed code rarely references it.

The second pattern: relying on the 1.2.3 transcript as if it cleared 1.2.5 too. At Level AA, it doesn't. The criteria are stacked, not substituted.

The third: confusing audio description with captions. A blind user needs the visuals narrated; a deaf user needs the audio captioned. The two tracks are independent. AI tools that add a <track kind="captions"> and call it done miss the description.

Failing examples

A <video> with captions but no description track:

<video controls>
  <source src="/launch-keynote.mp4" type="video/mp4" />
  <track kind="captions" src="/launch-keynote.en.vtt" srcLang="en" default />
</video>

A <track kind="descriptions"> pointing at a file that doesn't exist:

<video controls>
  <source src="/launch-keynote.mp4" type="video/mp4" />
  <track kind="descriptions" src="/launch-keynote.desc.vtt" srcLang="en" />
</video>

Passing examples

A described version of the video served as the default, with captions also available:

<video controls>
  <source src="/launch-keynote.described.mp4" type="video/mp4" />
  <track
    kind="captions"
    src="/launch-keynote.en.vtt"
    srcLang="en"
    label="English"
    default
  />
</video>

Alternative: an undescribed video plus a description WebVTT track that the browser reads aloud during pauses:

<video controls>
  <source src="/launch-keynote.mp4" type="video/mp4" />
  <track
    kind="descriptions"
    src="/launch-keynote.desc.en.vtt"
    srcLang="en"
    label="English descriptions"
  />
  <track
    kind="captions"
    src="/launch-keynote.en.vtt"
    srcLang="en"
    label="English"
    default
  />
</video>

Edge cases

Talking-head videos where nothing meaningful happens visually beyond the speaker's face often pass without an explicit description track — but the burden of proof is on the publisher.
Browser support for kind="descriptions" is uneven. The most reliable approach is a fully described version of the video as a separate source the user can switch to.
Extended audio description (1.2.7) pauses the video to fit longer narration. Use it when natural pauses are too short for adequate description.
Embed platforms (YouTube, Vimeo) support uploading audio description as a separate audio track. The feature exists; using it is a workflow decision, not a code one.
Recorded webinars that were live (1.2.4) now need audio description too once you publish the recording.

How Jeikin handles this

The scanner flags <video> elements that have no <track kind="descriptions"> and no alternate source for a described version, and maps each finding to WCAG 1.2.5. Confirming that the description track is accurate and complete is a manual review — the dashboard records the reviewer's confirmation, not an automated claim.