For each prerecorded time-based medium that has a synchronized audio description, is it relevant?
A blind user watches a product demonstration video: the audio description is present, it triggers correctly, but it just says "a person uses the software" without describing what appears on screen. Criterion 4.5 verifies the presence of audio description; criterion 4.6 verifies that it actually does its job.
A relevant audio description vocalizes all visual information that is not already conveyed by the main soundtrack, and that can be inserted into the natural silences of the video. This includes text overlaid on the screen (titles, speaker subtitles, interface names), significant actions, meaningful facial expressions, important scene changes. What someone who cannot see needs to know to follow the content.
The practical rule: if a blind user listened to only the audio track (original soundtrack + audio description), would they get exactly the same information as a sighted user? If the answer is no, the audio description is not relevant in the sense of RGAA.
2 tests to assess the relevance and completeness of the audio description
Relevance of audio description of video-only media
This test targets videos without original audio track (silent animations, demonstration videos without narration).
- Identify on the page all "video-only" videos (without sound) that have a synchronized audio description.
- For each one, view the video with audio description enabled and verify that:
- All text overlaid on the screen is read (titles, interface labels, speaker subtitles).
- Significant actions and visual information are described.
- The description fits into the silences without covering useful sound.
- Result: if each video meets these conditions, the test is validated. A single missing visual information causes the test to fail.
Relevance of audio description of synchronized media
This test targets synchronized media: videos with sound, recorded webinars, tutorials with narration and imagery.
- Identify on the page all synchronized media (video + audio) that have a synchronized audio description.
- For each one, compare what a sighted user sees with what a blind user hears (soundtrack + audio description). Verify that:
- Text overlaid on the screen is vocalized (slide titles, annotations, chart captions).
- Visual information not covered by narration is described.
- Audio description does not overlap with important spoken passages.
- Result: if each synchronized medium meets these conditions, the test is validated.
Examples
❌ Non-compliant : Vague audio description on an interface tutorial
<!-- SMIL audio description track -->
<smil xmlns="https://www.w3.org/2001/SMIL20/Language">
<body>
<par>
<video src="tuto-dashboard.mp4" region="video" title="Dashboard tutorial" />
<!-- Audio description: "The user navigates in the application" -->
<audio src="tuto-ad.mp3" begin="4s" title="Audio description" />
</par>
</body>
</smil>The audio description says "the user navigates in the application" without specifying: which menu is clicked, what text appears on screen, what numbers show in the table. A blind user does not know what has just been shown. The audio description exists, but it is hollow.
✅ Compliant : Precise audio description on the same tutorial
<!-- SMIL audio description track -->
<smil xmlns="https://www.w3.org/2001/SMIL20/Language">
<body>
<par>
<video src="tuto-dashboard.mp4" region="video" title="Dashboard tutorial" />
<!--
AD at 4s: "Click on the Reports menu at the top left.
A table appears: three columns,
Month, Sales, Target. January row:
12,400 euros, target 10,000 euros."
AD at 18s: "Export to PDF button, bottom right corner."
-->
<audio src="tuto-ad-detailed.mp3" begin="4s" title="Detailed audio description" />
</par>
</body>
</smil>Each visual action is described with precision: the name of the button clicked, the text content of the table displayed, the visible numbers. A blind user gets exactly the same information as a sighted user.
Tips and pitfalls
⚠️ Confusing presence with relevance
The most common mistake in audits: an audio description that describes movement ("the camera zooms in on the screen") rather than content ("the screen displays the error message: username or password incorrect"). Criterion 4.5 is satisfied, 4.6 is not. Always verify what is said, not just that something is said.
⚠️ Omitting text overlaid on the screen
RGAA explicitly cites embedded text as an element to vocalize. PowerPoint slide titles filmed, chart labels, names displayed under speakers in a conference: if these texts are not read by the audio description and are not present in the main soundtrack, the criterion fails. This is the non-compliance case most often found in audits of institutional websites.
⚠️ The main soundtrack already covers the information
If the narrator says "click the Send button, it's in the bottom right in blue", the audio description does not need to repeat this information. The criterion asks to vocalize what "can" be done "in the gaps of the main soundtrack". Do not duplicate: complement.
💡 Audit method: listening with eyes closed
Enable audio description, close your eyes, and listen. Note each time you feel that visual information has escaped you. Then compare with normal reading with eyes open. Each gap is a defect of relevance. This is the fastest and most reliable method, without specific tools.
⚠️ Extended audio description: outside the scope of criterion 4.6
When silences are not enough to insert all necessary descriptions, extended audio description is used (the video is paused). Criterion 4.6 also applies to this case: the relevance of the described content remains required, regardless of the synchronization mechanism used.
Frequently asked questions
How do I know if my audio description is "relevant" in the sense of RGAA?
Relevance is achieved when a blind user, listening to only the complete audio track (original sound + audio description), has all the information conveyed to the sighted user. Specifically: every text on screen, every significant action, every visual data that does not come from the voices present in the video must be described. If in doubt, have someone listen to the audio-only version who has not seen the video and ask them to describe what they understood from it.
Why can a video with complete narration still require an audio description according to RGAA?
If the narration covers all visual information (text, actions, displayed data), there is nothing to add and criterion 4.5 provides an exemption. But if visual elements provide information beyond what the voice says, an audio description remains necessary. Criterion 4.6 then applies: it must precisely describe these missing elements, not rephrase what is already said.
What visual elements must necessarily appear in an audio description that complies with RGAA?
No. RGAA asks to vocalize information "that can be vocalized in the gaps of the main soundtrack". Two criteria: the information must be useful for understanding the content, and the description must fit into an existing silence. A generic office background or decorative animation does not need to be described. A chart with specific numbers, yes.
What is the difference between RGAA criteria 4.6.1 and 4.6.2 for audio description?
4.6.1 targets "video-only" videos (without original sound: animations, silent screen captures). 4.6.2 targets synchronized media (video with sound: films, narrated tutorials, filmed conferences). The rule of relevance is identical in both cases; the distinction serves to properly target what type of content to audit.
How do I practically test the relevance of an audio description during an RGAA audit?
Enable audio description via the player or mechanism provided, then launch the video. Verify point by point: (1) are embedded texts read? (2) are key visual actions described? (3) are displayed data (numbers, names, interface states) vocalized? Document each omission with the corresponding timecode. Every omission of useful information is a defect of relevance.