For each prerecorded time-based medium that has a text transcript or a synchronized audio description, are they relevant (except in special cases)?
A deaf-blind person consults your page via a braille display. She cannot watch your video or listen to your podcast: her only entry point is the text transcript. If this document exists but reproduces an approximate summary rather than the actual content, she loses access to the information. Present but false is still a failure.
Criterion 4.2 does not verify the existence of the transcript or audio description—that is criterion 4.1's role. It verifies their quality. For audio-only media (podcast, interview, recorded conference), the transcript must faithfully reproduce the entire spoken content: identification of speakers, statements made, sounds carrying information (applause, alarm, meaningful silence). A transcript that summarizes or paraphrases fails, even if it is long.
For video without sound or synchronized media, the audio description must vocalize everything that the image communicates and that the soundtrack does not already say: on-screen text, important actions, location changes, names displayed in credits, data visible in a graph. It is inserted into the silences of the main soundtrack. A vague audio description such as "a person enters a room" is not relevant.
Auditing this criterion is a comparison task: read the transcript while following the media, or listen to the audio description while watching the video without sound. Any information present in the media and absent from the alternative constitutes a defect.
3 tests to assess the accuracy of the text transcript
Relevance of audio-only transcript
For each audio-only media (podcast, MP3 file, audio recording) that has a text transcript:
- Read the transcript while listening to the media.
- Verify that all statements are faithfully reproduced, with speaker identification.
- Verify that non-verbal sounds carrying information (meaningful music, alarm sound, determining atmosphere) are mentioned.
- If the transcript summarizes, paraphrases, or omits passages, the test fails.
The test is validated only if each found transcript is relevant.
Relevance of alternative to video-only media
For each video-only media (without audio track) that has a transcript or audio description:
- Identify which alternative is provided: text transcript, synchronized audio description, audio description of an alternative version, or audio-only version.
- Watch the video without sound, then compare with the retained alternative.
- Verify that all informative visual elements are covered: on-screen text, actions, identified characters, data displayed on screen.
- If a visual element carrying information is absent from the alternative, the test fails.
The test is validated if at least one of the available alternatives is relevant for each video.
Relevance of alternative to synchronized media
For each synchronized media (video with audio track) that has a transcript or audio description:
- Identify the available alternative: text transcript, synchronized audio description, or audio description of an alternative version.
- For the transcript: verify that it reproduces the audio content (dialogues, informative sounds) AND visual content (on-screen text, significant actions not described by dialogues).
- For the audio description: watch the video with the picture turned off and verify that the audio description compensates for everything that was visible but not said by the soundtrack.
- If visual or audio information is missing from the alternative, the test fails.
The test is validated if at least one of the alternatives is relevant for each synchronized media.
Examples
❌ Non-compliant : Podcast transcript reduced to an editorial summary
<audio src="podcast-accessibility-ep12.mp3" controls></audio>
<details>
<summary>Episode transcript</summary>
<p>In this episode, Julie and Marc discuss web accessibility
best practices and give advice to developers new to this field.
They cover ARIA, contrast, and keyboard navigation in particular.</p>
</details>This text is an editorial summary, not a transcript. It reproduces no actual statements: no dialogues, no identification of key moments, no sounds. A deaf-blind person retrieves no substantive information from the podcast. Present but false, criterion 4.2 fails, even though criterion 4.1 is satisfied because the transcript exists and is accessible.
✅ Compliant : Complete transcript with speaker identification
<audio src="podcast-accessibility-ep12.mp3" controls></audio>
<details>
<summary>Episode transcript</summary>
<div>
<p><strong>Julie:</strong> Today we're talking about ARIA attributes.
Marc, would you like to start with aria-label?</p>
<p><strong>Marc:</strong> Of course. aria-label allows you to give
an accessible name to an interactive element when no visible text
does so. An icon button without text must have
aria-label="Close" so a screen reader announces it correctly.</p>
<p><em>[Audience applause]</em></p>
<p><strong>Julie:</strong> Exactly. And what's the difference
with aria-labelledby?</p>
</div>
</details>Each speaker is identified, statements are reproduced faithfully, and non-verbal sounds carrying information (applause) are noted in brackets. A deaf-blind person accesses the same content as a sighted and hearing listener.
❌ Non-compliant : Vague audio description for a demonstration video
<!-- Excerpt from audiodescription.vtt file -->
WEBVTT
00:00:04.000 --> 00:00:07.000
A person works on a computer.
00:00:22.000 --> 00:00:25.000
She clicks on something and a result appears.The audio description does not specify what the person is doing or what result is displayed. If the video shows a contact form being filled in followed by the message "Your request has been sent. Response within 48 hours," this information is lost. A blind user does not understand the demonstration.
✅ Compliant : Precise audio description with reproduction of on-screen text
<!-- Excerpt from audiodescription.vtt file -->
WEBVTT
00:00:04.000 --> 00:00:07.000
Sophie, a front-end developer, fills in the "Subject" field
with the text "Quote request".
00:00:22.000 --> 00:00:25.000
She clicks "Send". A green banner appears:
"Your message has been sent successfully. Response within 48 hours."The audio description names the character, describes the precise action, and fully transcribes the on-screen text. A blind user receives the same information as a sighted viewer.
Tips and pitfalls
⚠️ Confusing editorial summary with transcript
This is the most frequent audit error. A transcript is not a general presentation of the topics addressed or a synthetic rewording. It reproduces the actual content of the media, statements made, identified speakers, meaningful sounds mentioned. A two-paragraph introductory text that "presents" the podcast is not a relevant transcript in the sense of criterion 4.2.
⚠️ Believing that captions replace the text transcript
Synchronized captions (criterion 4.3) and text transcript are two distinct alternatives that address different needs. Captions display during playback and are not accessible independently of the video player. The transcript is a separate document, accessible without launching the media, readable via a braille display, and indexable by search engines. One does not dispense with the other if both are required.
💡 Audit audio description with screen off
The most effective method: watch the video with the screen covered, listening only to the main soundtrack and audio description. If you understand all content without seeing the image, the audio description is relevant. Otherwise, each visual information not reproduced is a defect to document.
⚠️ Auto-generated transcripts are not reliable by default
Automatic transcription tools (AI-generated, YouTube auto-captions) produce often correct but unreliable results on proper names, technical terms, regional accents, and overlapping voices. An auto-generated transcript not reviewed and corrected regularly fails criterion 4.2. It is not the tool that is evaluated, but the quality of the final result.
⚠️ The special cases of criterion 4.1 apply here
Criterion 4.2 explicitly refers to the special cases of criterion 4.1. If a media is exempt from the obligation to provide a transcript or audio description (purely decorative content without information, content outside the legal scope according to the organization's status and publication date), the question of their relevance does not arise. Decorative media does not require an alternative, therefore no relevant alternative.
⚠️ For synchronized media, the transcript covers audio AND visual
For a video with sound, the text transcript must reproduce both audio content (dialogues, voice-over, meaningful sounds) and visual elements not covered by audio (on-screen text, important actions, on-screen data). A transcript that merely transcribes dialogues fails if the video contains visual information essential to understanding.
Frequently asked questions
What defines a relevant transcript according to RGAA criterion 4.2?
A transcript is relevant when it allows a person who cannot access the media to obtain all of its content. For audio: all statements reproduced faithfully, speakers identified, informative sounds mentioned. For video: all visual elements carrying information described. The question to ask yourself: does the transcript completely replace the media for someone who cannot access it? If the answer is no, it is not relevant.
How to handle background music and ambient noise in RGAA transcript?
Only if it carries information. Neutral background music does not need to be mentioned. However, if the music changes abruptly to signal a moment of tension in a documentary, or if an off-screen noise (gunshot, alarm, telephone ring) is relevant to understanding the scene, it must appear in the transcript. The practical rule: transcribe what a hearing listener would get from this sound to understand the content.
How to audit the relevance of an audio description in the RGAA framework?
Proceed in two stages. First, watch the video normally and note each visual element carrying information: on-screen text, displayed names, important actions, visible data on screen. Next, listen only to the soundtrack with the audio description without watching the screen. Each item on your list must be covered by the audio. What is missing is a defect to report.
How does RGAA criterion 4.2 apply to media without an existing transcript?
It does not. If no transcript or audio description is present when required, it is criterion 4.1 that fails. Criterion 4.2 is evaluated only on alternatives already provided. Both criteria can therefore fail simultaneously, but on different media: 4.1 on those that have no alternative, 4.2 on those whose alternative exists but is insufficient.
What relevance requirements apply to extended audio descriptions per RGAA?
Yes. Criterion 4.2 concerns the relevance of synchronized audio description, whether standard or extended. Extended audio description (which pauses the video to insert long descriptions) follows the same rules: all visual information must be vocalized. The distinction between standard and extended falls under criteria 4.13 and 4.14, which verify the presence of this extended audio description, not its quality.