WCAG 2.1 — what's required
Two success criteria cover video captions: SC 1.2.2 (Captions, prerecorded — Level A) requires captions on all prerecorded video with audio, and SC 1.2.4 (Captions, live — Level AA) extends the requirement to live video. AA is the practical baseline for most public sites; AAA adds extended audio descriptions and sign language.
ADA — the US legal layer
The ADA itself doesn't name WCAG, but courts and the DOJ have repeatedly cited WCAG 2.1 AA as the standard for digital accessibility. If you're a public-facing US business with a website, the practical requirement is: captions on every video you publish.
Section 508 — federal contracts
Section 508 explicitly requires WCAG 2.0 Level AA conformance for federal agencies and contractors. Video captions are mandatory; missing captions can disqualify a deliverable.
What 'compliant' captions look like
- Synchronized to the audio (no drift, no missed lines).
- Equivalent — every spoken word, plus speaker identification when not visually obvious.
- SDH-style — non-speech audio cues like [music] or (door slams) when relevant.
- Toggleable — the user must be able to turn captions on. Closed captions or a separate caption track preferred.
- Readable — sufficient contrast, no obscured by chrome.
Are auto-captions good enough?
Generally no. WCAG specifies captions must be 'equivalent' to the audio. Auto-caption error rates of 5–15% on clean English (higher on accents, music, or jargon) typically fall short. The accepted standard is human-reviewed AI captions: generate with Whisper or similar, then proofread for the meaningful errors before publishing.
A practical compliance workflow
- 01Generate captions with an AI tool (Whisper, SoCaptions, AssemblyAI).
- 02Proofread for proper nouns, technical terms, numbers, and names.
- 03Add speaker labels and non-speech cues for SDH.
- 04Embed as closed captions (HTML5 <track>, YouTube CC, platform native) so users can toggle.
- 05For social where toggleable captions are unreliable, also burn captions into the video as belt-and-braces.
FAQ
What does WCAG say about video captions?
WCAG 2.1 Success Criterion 1.2.2 (Captions, prerecorded — Level A) requires synchronized closed captions for all prerecorded video that has audio. SC 1.2.4 (Captions, live — Level AA) extends the requirement to live streams. Both require the captions to be 'equivalent' to the audio, meaning every spoken word plus significant non-speech audio like music cues and sound effects.
Are YouTube auto-captions WCAG compliant?
No — not without review. WCAG 1.2.2 requires captions to be 'equivalent' to the audio. Auto-generated captions with 5–15% word error rate (higher for technical terms and accents) fail equivalence audits. The compliant approach is AI-generated captions reviewed and corrected by a human before publication. YouTube's auto-captions are fine as a first pass; they're not acceptable as the final accessibility deliverable.
Do burned-in captions meet WCAG requirements?
No — WCAG requires captions that can be toggled on and off by the viewer and accessed by assistive technologies. Burned-in captions are pixels in the video frame; a screen reader cannot parse them and the viewer cannot disable them. Deliver a proper SRT or VTT track for WCAG compliance. Burned-in captions on social video are fine as a supplement, but they don't satisfy the requirement alone.
What is the difference between WCAG Level A and Level AA for captions?
Level A (SC 1.2.2) requires captions on all prerecorded video with audio. Level AA (SC 1.2.4) additionally requires captions on live video. Most enterprise procurement and ADA compliance targets WCAG 2.1 AA, which means both prerecorded and live video need captions. Level AAA adds extended audio descriptions, but most organizations don't target AAA.