WCAG 1.2.4 Captions (Live): Real-Time Access Is a Right

No live captions means no access — not later, not approximately, but in the moment. A recording with captions added afterward doesn't restore what was missed during a live town hall, a breaking news broadcast, or a webinar Q&A. WCAG Success Criterion 1.2.4 (opens in new window) exists because real-time exclusion is its own category of harm, distinct from the barriers addressed by its companion criterion for pre-recorded content.

This is a Level AA requirement, which means it sits at the core of what most organizations are legally obligated to meet — not a stretch goal, not aspirational. For any organization broadcasting live synchronized media, this is the baseline.

The Requirement

WCAG 1.2.4 (opens in new window) requires captions for all live audio content in synchronized media. The W3C's intent language is precise: captions must include not just dialogue, but speaker identification and notation of significant sound effects. A transcript of words spoken is not a caption. A scrolling text feed that arrives 45 seconds late is not a caption. Synchronization is the requirement, not just text.

One important scope clarification in the official Understanding document: this criterion was designed for broadcast-style synchronized media. It is not intended to require that every two-way multimedia call between individuals must be captioned regardless of user needs. Responsibility in those contexts falls to content providers or host callers, not the platform application itself. That distinction matters for teams building video conferencing tools versus teams hosting public-facing live events.

Requirement	Detail
Level	AA
Applies to	Live synchronized media with audio
Captions must include	Dialogue, speaker ID, significant sound effects
Synchronization	Real-time — not post-event transcripts
Scope exception	Two-way individual calls (not broadcast/public events)
Primary authority	W3C WCAG 2.2 Understanding 1.2.4 (opens in new window)
Governing rule (Title II)	28 CFR Part 35 (opens in new window)

Who Is Excluded Without It

Approximately 15% of American adults report some degree of hearing difficulty, according to the National Institute on Deafness and Other Communication Disorders (opens in new window). For Deaf users who rely on ASL as a primary language, live captions are not a supplement — they are the access point. For hard-of-hearing users in noisy environments, on mobile devices, or with audio processing differences, live captions are often the difference between participation and exclusion.

The harm is time-bound in a way that recorded content is not. A person who misses a live earnings call, a city council meeting, a university lecture, or a congressional hearing because captions weren't provided cannot simply "catch up" in a way that restores equal participation. The Q&A is over. The vote has been taken. The professor has moved on. This is why community-centered accessibility analysis consistently identifies live captioning gaps as high-priority barriers — the exclusion is immediate, complete, and often irreversible within the context of that event.

Users with cognitive disabilities also benefit from live captions. Reading along while listening supports comprehension for people with auditory processing differences, attention-related disabilities, and some learning disabilities. The benefit extends well beyond the Deaf and hard-of-hearing communities typically cited in the criterion's intent language.

How to Meet It

The W3C identifies five sufficient techniques for satisfying 1.2.4:

• G9 (opens in new window) — Creating captions for live synchronized media. This is the general technique, covering any method that produces real-time synchronized captions.
• G93 (opens in new window) — Providing open (burned-in) captions. These are always visible, requiring no user action to enable.
• G87 (opens in new window) — Providing closed captions. User-toggled, typically via a CC button in the media player.
• SM11 — Using SMIL 1.0 to provide captions for synchronized media.
• SM12 — Using SMIL 2.0 to provide captions for synchronized media.

In practice, most organizations delivering live captions today use one of three approaches:

Automated Speech Recognition (ASR) — Platforms like YouTube Live, Zoom, and Microsoft Teams offer real-time auto-captions. Accuracy rates vary significantly by speaker accent, technical vocabulary, audio quality, and background noise. ASR captions can satisfy the technical requirement but often fail the spirit of it when accuracy drops below usability thresholds. This is a known gap that automated testing cannot catch — a caption stream can exist and still be functionally inaccessible.

Communication Access Realtime Translation (CART) — A trained human stenographer produces near-verbatim captions in real time, typically achieving 98%+ accuracy. CART is the gold standard for high-stakes events: legal proceedings, medical appointments, academic lectures, government hearings. Cost is real — professional CART services typically run $100–$200 per hour — but the accuracy differential is significant.

Hybrid approaches — Some organizations use ASR as the base layer with human editors correcting the stream in near-real-time. Emerging AI-assisted captioning tools are attempting to close the accuracy gap while reducing cost.

Common failures to watch for:

• Caption streams that launch 30+ seconds after audio begins
• Auto-captions that drop speaker identification entirely
• Live events where captions are "available on request" but require advance notice — this is not equivalent access
• Platforms that provide captions in recordings but disable them for live streams

Applying This

For development and QA teams, live captioning presents a specific challenge: automated testing tools cannot verify caption accuracy or synchronization quality. A tool can confirm that a caption track exists. It cannot tell you whether the captions are 45 seconds behind, whether speaker identification is present, or whether the accuracy rate is sufficient for comprehension.

Practical steps for teams:

In design and procurement: When evaluating live streaming platforms, require vendors to document their captioning architecture. Ask specifically: Is ASR accuracy tested against diverse speaker accents? Can CART integration be enabled? Is there a latency SLA for the caption stream?

In QA: Manual testing of live caption quality should be part of any pre-launch checklist for live event infrastructure. Test with actual diverse audio inputs — not just a single speaker in a quiet room. Review latency, accuracy on technical terminology, and speaker identification.

In operations: For high-stakes events (public meetings, academic instruction, healthcare communications), default to CART rather than ASR. The cost of inaccessible captions — in both human terms and legal exposure under Title II and Title III (opens in new window) — exceeds the cost differential.

In procurement contracts: If your organization hosts live events through a third-party platform, ensure your contract specifies captioning requirements and accuracy standards. Platform defaults are often insufficient.

Organizations navigating multiple compliance frameworks — WCAG 2.2, Section 508, EN 301 549 — will find that live captioning requirements are consistent across all three. This is one area where standards alignment actually simplifies the compliance picture rather than complicating it.

CORS Perspective

From a community-centered lens, live captioning gaps represent a particularly acute form of exclusion because they are time-sensitive and often invisible to the organizations creating them. A city government that adds captions to archived meeting recordings but runs live council sessions without CART is not providing equivalent access — it's providing delayed access to decisions already made. The operational investment in CART services or vetted ASR platforms is real, but it's also predictable and plannable, unlike litigation costs. Strategically, organizations that build live captioning into their standard event infrastructure — rather than treating it as a special accommodation — reduce both their legal exposure and the administrative burden of fielding individual accommodation requests. The data suggests this is a moment when getting the infrastructure right actually pays forward.