WCAG Live Captions: What 1.2.4 Requires and How to Meet It

WCAG requires captions for all live audio in synchronized media, such as a webcast, a livestreamed service or a virtual meeting with video. That is Success Criterion 1.2.4, a Level AA requirement. To meet it, you need real-time text that stays in sync with the speaker. A transcript posted after the event does not count. How accurate that text has to be, and who produces it, is where most organizations get stuck. This guide covers both.

TL;DR

• WCAG 1.2.4 (Level AA) requires synchronized captions for live audio content. A post-event transcript does not satisfy it.
• Level AA is the usual target in US accessibility rules and policies, so live video with meaningful audio generally needs live captions.
• Accuracy matters as much as presence. W3C warns that automatic captions usually need significant editing to meet user needs.
• Preparation drives quality: load names and specialized terms before you go live, and plan early for any human captioner.
• Live Caption AI is an AI captioning tool. It can be one part of your plan. It does not make you compliant on its own.

Table of Contents

• What WCAG 1.2.4 actually says
• Where WCAG, the ADA and Section 508 meet
• The real question: how accurate is accurate enough?
• Plan early: the most common mistake
• How a QR code approach works for live events
• Where AI captions fit in a WCAG and ADA plan
• Our perspective: preparation beats the stream
• Meeting WCAG 1.2.4: a checklist
• Costs and plans
• FAQ
• Sources

What WCAG 1.2.4 actually says

The criterion is one line: captions are provided for all live audio content in synchronized media. The W3C’s own explanation of its intent is to let people who are deaf or hard of hearing watch real-time presentations. Three terms carry the weight.

• Live. The audio is happening now. Recorded material falls under a different criterion, 1.2.2 (Captions, Prerecorded).
• Synchronized media. Audio paired with video or other time-based visuals. A livestream of a Sunday service qualifies. So does a webinar with slides and a speaker on camera.
• Captions. Not just words. The W3C describes captions as text for speech and for the non-speech audio a viewer needs to understand the content: sound effects, music, laughter and who is speaking.

That last point catches people out. A caption stream that gets every spoken word right but never says who is talking, or ignores a meaningful sound, is still incomplete. Section508.gov makes the same point: good captions identify speakers, note significant sounds and convey tone where it changes meaning.

Captions are not transcripts

A transcript is a document. A caption is real-time, synchronized text. Criterion 1.2.4 asks for the second. If you only publish a transcript after the event, you have done something useful, but you have not met 1.2.4 for the live session.

There is one nuance. Audio-only content, like a live podcast with no video, is treated differently under WCAG’s time-based media rules. A transcript is the usual alternative there, and it still helps listeners. Once there is video with meaningful audio, live captions come into play at Level AA.

Where WCAG, the ADA and Section 508 meet

WCAG is a technical standard. The law that makes it matter in the US comes from elsewhere, and the pieces overlap.

The ADA asks for effective communication

The ADA does not name a product. ADA.gov guidance on effective communication says the aid or service you provide depends on the complexity of the communication, the needs of the people involved and the setting. For people who are deaf or hard of hearing, that can include real-time captioning, also called CART (communication access realtime translation).

The test is outcome-based. The communication has to be as effective for a person with a disability as it is for everyone else. The same guidance expects aids to be qualified, meaning they can convey information accurately and impartially. That is the standard to hold any captioning method to, human or AI.

Section 508 covers federal and federally funded work

If you produce video for a federal agency, or your work is funded in a way that carries Section 508 obligations, the rules apply directly. Section508.gov’s guidance on captions and transcripts lays out the technical standards, legal requirements and best practices. Its page on video and other synchronized media explains that captions display spoken dialogue and other sounds on screens and projected video.

So a question like “what are the best 508 compliant captions?” has a plain answer. No tool is 508 compliant in the abstract. Compliance is a property of what you deliver: the captions are synchronized, complete and accurate enough for the people relying on them. Choose a method that can reach that bar for your content, then check the output.

The real question: how accurate is accurate enough?

WCAG 1.2.4 does not set a percentage. It does not need to. The ADA’s effective communication test and the W3C’s guidance both point at the same thing: the captions must work for the person reading them.

The W3C is direct about automatic captions. Its captioning guidance notes that live captions are usually produced by professional real-time captioners or CART providers, and that automatically generated captions usually need significant editing to meet user needs. AI captions that mangle names, drug terms, scripture references or financial figures are likely to fall short.

This is not a reason to avoid AI. It is a reason to be honest about where it fits. Think in terms of risk:

• Lower stakes, broad audience. A church service, a town hall, a conference session, a class lecture. AI captions on attendees’ own phones add access that would otherwise not exist, especially when the budget for a human captioner is not there.
• Higher stakes, specific individual. A court proceeding, a medical consultation, a disciplinary hearing, a situation where a person has asked for a particular accommodation. Here, the ADA’s effective communication test may call for a human captioner or another verified method. Ask the person what works for them.

Many organizations assume that switching on an automated tool settles the question. It does not. If the output is inaccurate or hinders communication, the obligation is not met.

What drives caption quality in practice

Technology is only part of it. Four things matter more than most buyers expect:

1. Audio quality. A clean microphone close to the speaker beats any software setting. Test the room before the event.
2. Vocabulary preparation. Names, places and jargon are where captions break. Load them in advance.
3. Speaker habits. Steady pace, one speaker at a time, and repeating audience questions into the mic.
4. A person who checks. Someone from your team should watch the captions during the session and fix what they can afterward.

Plan early: the most common mistake

The Department of Justice’s guidance and the experts around it keep returning to timing. Organizations wait until the last minute to arrange live captioning. Professional human captioners can need weeks of notice, and good ones book up. Build accessibility into event planning from the start, and tell attendees in the invitation how to request services.

A simple timeline helps:

For a longer walk-through of event logistics, see our live captioning for events planning guide.

How a QR code approach works for live events

Traditional live captioning needs a captioner, a feed and a screen. For a room full of people, one screen is a compromise: those sitting at the back or at an angle cannot read it, and someone who needs captions may not want to sit front and center.

Phone-based captions solve a different problem. Each person reads on their own device, at their own text size, from wherever they sit. With Live Caption AI the flow is short:

1. The speaker opens Live Caption AI on a phone or laptop and starts a session.
2. A QR code appears. Put it on a slide, a sign or the program.
3. Attendees scan it and read live captions in their browser. No app to install. No account to create.
4. When the session ends, you save the transcript.

No hardware to rent. No stenographer to book. Session audio is never stored.

Specialized vocabulary

Generic speech-to-text stumbles on proper nouns. Live Caption AI offers a Medical model, a Worship & Ministry model and a Finance & Earnings model. You can also add Custom Terms, up to 100 names, places and terms typed in before you go live. That is the preparation step in practice: the pastor’s guest speaker, the name of the venue, the product line, the board members.

Two limits to know. There is no legal or technical vocabulary model. And Custom Terms are typed in by the speaker, not uploaded as a glossary file.

Translation

For multilingual audiences, Live Caption AI translates into one of 29 languages per session. The speaker picks one source language and one target language, and viewers can switch between the original and the translation. It is one target language per session. Translation is text only, with no translated audio.

Transcripts and summaries

You can save and export transcripts as TXT, SRT, JSON or PDF. AI summaries are included. A transcript is a useful companion to live captions, and it is what you would post after the event for anyone who missed it. Remember that it complements 1.2.4 and does not replace it.

Where AI captions fit in a WCAG and ADA plan

Be straightforward with your board, your counsel and your attendees about what an AI tool is. Live Caption AI is an AI captioning tool. It is not a certified service, and it is not human-reviewed. It does not make an organization ADA compliant on its own. An organization can use it as one part of its ADA effective-communication plan.

A workable plan has layers:

• Default access for everyone. AI captions via QR code at every event, so access does not depend on someone asking in advance.
• A process for requests. A line in every invitation and on your website, with a named contact and a response time.
• Escalation for specific needs. If a person tells you AI captions do not work for them, or the setting is high stakes, arrange a human captioner or another aid they find effective.
• Quality habits. Term lists, mic checks, someone monitoring, and review of transcripts afterward.
• Records. Keep a short log of what you provided and what you changed. It shows good faith and helps you improve.

The responsibility for whether captions are good enough for a given participant stays with the host. A tool can lower the cost and effort of getting there. It cannot take on the duty.

Our perspective: preparation beats the stream

We build Live Caption AI, and we have a firm view on this topic. Most organizations that ask about WCAG live captions are not trying to dodge anything. They are a church with a volunteer tech team, a nonprofit with a small events budget, or an AV company that has a client in the room with a multilingual crowd. They need something that works this week, at a cost they can carry.

Here is what we have learned shapes results.

Captions fail on names. A generic engine hears “Habakkuk” or “metoprolol” or a company’s new product name and guesses. That is why we built the specialized models and Custom Terms. Ten minutes of typing your terms in before you go live changes what attendees read.

Access should not require asking. Many people will not announce that they need captions. If the QR code is on the screen and the program, anyone can use them, including people who are hard of hearing but never asked, and people who simply follow better in text.

We are also clear about limits. We are not a certified or human-reviewed service, and we do not claim ADA certification. If you have an attendee whose situation calls for a human captioner, book one. If you want captions for everyone else in the room, at a flat monthly price, that is where we fit.

— Ryan

Meeting WCAG 1.2.4: a checklist

Use this before any live session with video:

• Is there real-time text for all meaningful audio, not just a transcript promised later?
• Can attendees open captions on their own device without installing anything or signing up?
• Is the QR code visible on screen, in print and in the invitation?
• Have you loaded speaker names, place names and specialist terms?
• Does the text identify who is speaking when it matters?
• Is someone assigned to watch the captions and flag problems?
• Does the invitation tell people how to request other accommodations?
• Do you have a fallback if a person needs a human captioner?
• Will you export, review and post a corrected transcript afterward?

If you stream the event on video, also check that your streaming page itself shows captions or links to the caption view, so remote viewers have the same access as the room.

Costs and plans

Human captioning is a skilled professional service, and it is priced that way. The budget gap is why many smaller organizations skip captions entirely. That is the outcome we want to prevent.

Live Caption AI has a Free plan with basic on-device captions. The Professional plan is $19.99 a month with unlimited captioning. The Business plan starts from $199 a month. If you need translation beyond what your plan includes, one-off credit packs are available: 30 credits for $6.99, 60 credits for $11.99 and 120 credits for $19.99. Current details are on the pricing page.

Thinking about a live service, a class or an event series? Explore Live Caption AI and set up a first session with your own term list.

FAQ

Does WCAG require live captions?

Yes, at Level AA. Success Criterion 1.2.4 requires captions for all live audio content in synchronized media. Level A alone does not include it, so organizations aiming for AA, the common target, need live captions for live video with meaningful audio.

Are AI-generated live captions enough for WCAG?

They can satisfy the requirement to provide captions, but accuracy is the catch. The W3C notes that automatic captions usually need significant editing to meet user needs. Use AI captions with preparation and monitoring, and be ready to arrange a human captioner when a person or setting calls for it.

Is a transcript a substitute for live captions?

No. A transcript is a document, and 1.2.4 asks for real-time, synchronized text. Post a transcript after the event as a bonus for people who missed it or want to review. Do not rely on it alone for the live session.

What are the best 508 compliant captions?

Section 508 does not certify tools. It looks at what you deliver. The best option is the one that produces synchronized, complete and accurate captions for your audience and content. For high-stakes settings that often means a human captioner. For broad audiences, AI captions with prepared vocabulary can be part of your plan.

Do attendees need to install an app to read Live Caption AI captions?

No. Attendees scan a QR code and read captions in their phone’s browser. There is no app to install and no account to create.

Sources

• Understanding Success Criterion 1.2.4: Captions (Live), W3C WAI
• Captions/Subtitles, W3C Web Accessibility Initiative
• ADA Requirements: Effective Communication, ADA.gov
• Captions and Transcripts, Section508.gov
• Video and Other Synchronized Media, Section508.gov

Recommended

• ADA Compliant Event Captions Using Cloud Live Transcription via QR