Does each prerecorded synchronized time-based medium have, where necessary, synchronized captions (except in special cases)?
A deaf person opens your online training video. She sees the slides scroll by, the presenter moving their lips. Without captions, the speech disappears entirely. No captions, no access.
Two valid solutions exist to satisfy this criterion. Either the captions are directly integrated into the media — embedded in the video or via a separate activatable track. Or you provide an alternative version with synchronized captions, accessible from a link or button placed right next to the player. Warning: this alternative version cannot be buried three clicks deep in navigation.
In native HTML, the recommended technique is the <track> tag with the kind="captions" attribute. It points to a WebVTT file containing the captions and their temporal synchronization. Complementary criterion 4.4 will then verify that the content of these captions is truly relevant — transcription of dialogue and informative sounds.
Embedded captions or separate track? Both comply with criterion 4.3. The difference: embedded captions are always visible, with no user control possible. A separate track allows the user to activate or deactivate captions according to their needs — criterion 4.11 then imposes a control mechanism in the player.
2 tests to verify the correct synchronization of subtitles
Presence of synchronized captions on video
- List all pre-recorded video media with soundtrack in the document (
<video>elements,<object>, embedded players). - For each one, verify that at least one of the following two conditions is met:
- Synchronized captions are present in or with the player (embedded in the video, or separate activatable track);
- An adjacent link or button points to an alternative version of the video that contains synchronized captions.
- If neither condition is met for at least one media, the test fails.
kind="captions" attribute on <track> element
- Identify videos that use a
<track>tag for their captions. - Inspect each
<track>tag in the source code (developer tools, Elements tab). - Verify that the
kindattribute is indeed equal to"captions"— not"subtitles", not"descriptions", not an absent value. - If all caption
<track>tags havekind="captions", the test is validated.
Examples
❌ Non-compliant : Conference video without caption track
<video controls poster="conference.jpg">
<source src="conference.mp4" type="video/mp4">
<source src="conference.webm" type="video/webm">
</video>No <track> tag and no link to an alternative version. A deaf or hard-of-hearing person receives no information about the audio content of the conference. Criterion 4.3 is violated.
❌ Non-compliant : Video with <track kind="subtitles"> track
<video controls poster="conference.jpg">
<source src="conference.mp4" type="video/mp4">
<track kind="subtitles" label="Français" src="conference-fr.vtt" srclang="fr">
</video>The value kind="subtitles" designates a translation for non-French speakers, not accessibility captions. Test 4.3.2 fails despite the presence of a VTT file. The correct value is kind="captions".
✅ Compliant : Video with correctly marked accessibility captions
<video controls poster="conference.jpg">
<source src="conference.mp4" type="video/mp4">
<source src="conference.webm" type="video/webm">
<track kind="captions" label="Sous-titres français" src="conference-fr.vtt" srclang="fr" default>
</video>The <track kind="captions"> tag points to a synchronized WebVTT file. The default attribute activates captions on launch. Deaf and hard-of-hearing users access the full audio content without additional action.
Tips and pitfalls
⚠️ kind="subtitles" fails test 4.3.2 even with identical content
In RGAA audit, this is one of the most frequent errors on this criterion. "subtitles" designates a translation for an audience that understands images but not the language. "captions" designates accessibility captions — which include dialogue, important sound effects ([telephone ringing], [applause]) and musical information. RGAA explicitly requires kind="captions". An identical VTT file with kind="subtitles" constitutes a failure.
⚠️ Media offered as an alternative to textual content on the page
If a video is offered as an alternative to text already present on the same page, and clearly identified as such, it is exempt from the caption requirement. The mention must be explicit: "This video illustrates the written tutorial above". The exemption falls as soon as the video contains information absent from the reference text. In case of doubt, add captions — compliance review will compare the two contents.
⚠️ Embedded captions comply without control button
Captions graveled directly into the video's pixels (open captions) satisfy criterion 4.3 without any activation mechanism being required. They are always visible. The downside: they cannot be disabled or customized by the user. For separate tracks (closed captions), an activation button in the player becomes mandatory — criterion 4.11 imposes it.
💡 Test the track with a screen reader, not just visually
A caption displayed on screen is not necessarily accessible to NVDA or VoiceOver. Some video players do not expose the <track> track to the browser's accessibility API. Test with NVDA on Firefox: activate captions, launch the video, and verify that captions are announced in real time. If they appear visually but remain silent for the screen reader, it is the video player itself that is non-compliant (criterion 4.11).
⚠️ Content published before the legal deadline of September 23, 2020
For public sector organizations, content published before September 23, 2020 may benefit from a temporary exemption under certain conditions, arising from Directive 2016/2102 transposed into French law. This waiver does not apply to new content or substantial updates. A video that is reposted or republished after this date loses the benefit of the exemption.
Frequently asked questions
What is the difference between kind="captions" and kind="subtitles" for RGAA captions?
captions designates accessibility captions: they transcribe dialogue AND sounds that carry information — sound effects, music, significant silences. subtitles designates a translation of the spoken language for people who understand images but not the language. RGAA and WCAG H95 technique require kind="captions" to satisfy criterion 4.3.2. Using kind="subtitles" with identical content constitutes a failure on the test.
How do you audit RGAA criterion 4.3 on synchronized captions?
Open developer tools and inspect each <video> — look for a <track kind="captions">. For embedded players (YouTube, Vimeo), manually verify that captions are available and activatable from the player. If the player does not support <track>, look for a link or button adjacent to an alternative version with captions. Activate captions, launch the video and confirm synchronization. Then test with NVDA to validate the accessibility of the track to screen readers.
Why doesn't a text transcript below the video validate RGAA criterion 4.3?
No. The transcript answers criteria 4.1 and 4.2. Criterion 4.3 requires synchronized captions — displayed in real time during playback, aligned with speech. A text box below the video does not replace synchronized captions. The two are complementary and can coexist: the transcript also helps search engines and users who prefer to read.
How do you manage RGAA criterion 4.3 for a YouTube video without HTML access?
On YouTube, activate captions in the video manager — upload a manual VTT file rather than relying on automatic captions, whose quality is often insufficient to meet criterion 4.4. Integration via <iframe> does not allow adding a <track>, but YouTube captions remain accessible directly in the embedded player. Verify that the captions option is visible and functional in the iframe.
What must a WebVTT file contain to comply with RGAA?
The file must transcribe the entire dialogue and all sounds that carry information: significant sound effects ([ringing], [fire alarm]), musical indicators ([dramatic music]), identification of speakers if multiple people speak. Omitting sound effects and transcribing only dialogue constitutes a failure on criterion 4.4 on the relevance of captions — distinct from criterion 4.3 which verifies their presence.