What Is Closed Captioning? Definition, Formats, Legal Requirements (2026)

TL;DR — quick definition

Closed captionsare toggleable on-screen text representing dialogue plus non-speech audio (music, sound effects, speaker labels) for viewers who cannot hear. The “closed” part means the viewer can turn them on or off — as opposed to open captions, which are permanently visible. Not the same as subtitles, which show dialogue only and are historically for viewers who can hear but don’t understand the language.

Closed captions vs open captions vs subtitles vs SDH

Four related terms that get used interchangeably in casual speech but have distinct meanings.

The regional inconsistency

The captions-vs-subtitles distinction above is standard in North America. In the UK and much of Europe, “subtitles” is the umbrella term for both same- language and translated timed text, and “closed captions” is less common as a category — the BBC, for example, refers to accessibility-focused text as “subtitles for the hard of hearing.”

For technical work: the file format (SRT, WebVTT) doesn’t care what you call the content. The distinction matters for legal compliance and platform metadata (Netflix, for example, expects a specific SDH track).

Who is legally required to provide closed captions

Not legal advice — cite the linked sources for specific compliance questions.

Private commercial video content in most jurisdictions isn’t universally legally required to have captions, but WCAG 2.1 is the widely-adopted accessibility standard for digital content. Meeting it protects against most accessibility lawsuits and satisfies government/enterprise procurement requirements.

How closed captions are produced

Three methods, with tradeoffs between accuracy, cost, and turnaround.

Automatic caption tools

Most modern captioning tools use OpenAI’s Whisper or a similar transformer-based ASR model. Accuracy on clean English studio audio hits 95–98%. Real-world accuracy varies: expect 80–90% on Zoom calls, 65–85% on phone audio, 50–75% on noisy or heavily accented content. Full breakdown on How Accurate Is Whisper.

Human captioning

Professional captioners hit 99%+ accuracy and handle edge cases (proper nouns, technical jargon, non-speech markers) that AI misses. Services include Rev human-reviewed, Trint human tier, and specialist captioning houses. Turnaround is 24–72 hours; some offer rush pricing.

Hybrid (AI + human)

AI generates a first draft, human reviewers correct errors and add non-speech markers. Combines AI cost with human accuracy. Most cost-effective path to publication-quality captions at scale.

Caption file formats explained

Which format should I use?

• Web video: WebVTT (.vtt) — W3C standard, HTML5 native
• YouTube / Vimeo: SRT (.srt) — universal, simplest
• Streaming platform delivery (Netflix, Prime): TTML/DFXP — per platform spec
• US broadcast: SCC — EIA-608 encoding
• Podcasting 2.0: WebVTT — see Podcast Transcription

For SRT specifics (line-break rules, character limits, timing), see SRT Generator.

How closed captions are displayed

Broadcast TV

Broadcast captions use EIA-608 (older, analog-derived) and EIA-708(digital HDTV standard). CC data rides in the video signal, decoded by the TV or set-top box. Home TVs have the CC toggle in the accessibility settings; captions render in the TV’s built-in font at platform defaults.

Web video (HTML5)

HTML5 video elements use the <track> tag to reference a WebVTT file:

<video controls>
  <source src="video.mp4" type="video/mp4">
  <track kind="captions" src="captions.vtt" srclang="en" label="English" default>
</video>

Browsers render the captions with their native controls; styling is limited (positioning + basic color).

Streaming platforms

Netflix, Prime Video, Disney+, and HBO all use their own player implementations of TTML/DFXP tracks. Users see the CC button in the player; captions render with platform-specific styling (usually customizable in the user settings).

Social media (burned-in)

Instagram Reels, TikTok, YouTube Shorts, and Facebook typically burn captions in at upload — the caption text is rendered into the video frames before playback. Cannot be toggled off. See Caption Generator for the burn-in workflow with FFmpeg force_style.

Common confusions

“Captions” and “subtitles” are synonyms

Not in North American accessibility contexts. Captions include non-speech audio (music, sound effects, speaker ID); subtitles show dialogue only. In casual UK/EU usage, “subtitles” often covers both.

“CC always means AI-generated”

No.CC refers to the ability to toggle captions on/off — not how they were produced. Both human-authored and AI-generated captions can be “closed.” The distinction is presentation, not authorship.

“CC and SDH are the same thing”

Mostly, but not exactly. Both include non-speech audio for deaf/HoH viewers. Historical distinction: CC was designed for broadcast TV with specific technical constraints (character limits per line, positioning rules); SDH was streaming/DVD-era term for the same content in modern formats. Today the terms are used interchangeably by most platforms.

“SRT is the official caption format”

No — SRT is the most universally-supported informal format. The official formats are: SCC/EIA-608/708 for US broadcast, WebVTT for the web, TTML/DFXP for streaming. SRT is a de-facto standard for interoperability but not a formal broadcast or web spec.

“Auto-captions meet accessibility requirements”

Usually no.WCAG 2.1 requires captions that are “equivalent to spoken dialogue and equivalent information conveyed through sound.” Auto-captions typically hit 80–90% accuracy on clean English audio and skip non-speech markers entirely — not legally sufficient for accessibility compliance in most contexts. Human review is usually required.