Closed Captions vs Subtitles: The Difference and the Rules

The difference

The W3C's accessibility pages define captions as a text version of the speech and non-speech audio information needed to understand the content. They use captions for text in the same language as the spoken audio, and subtitles for spoken audio translated into another language. Captions are an accessibility feature; foreign-language subtitles generally are not, because they leave out the sounds a hearing viewer already hears.

The two words are often used loosely, and a menu may list both under one label. What matters is what the track contains.

Open and closed captions

Closed captions can be turned on and off by the viewer. Open captions are burned into the picture and are always visible. Most captions are closed. Open captions are part of the image, so a viewer cannot turn them off, restyle them or swap them for another language, and they can cover parts of the frame. For the vertical-video case, the guide to vertical video safe zones shows where on-screen text gets covered by app buttons.

What a good caption contains

The U.S. television rules in 47 CFR 79.1 set four quality standards. In short:

• Accuracy. Captions match the spoken words in the order spoken, without paraphrasing except where time makes it necessary, with correct spelling and punctuation. They also give nonverbal information that is not observable, such as the identity of speakers, the existence of music, sound effects and audience reaction.
• Synchronicity. Captions appear when the speech or sound begins, end about when it ends, and stay up long enough to be read.
• Completeness. They run from the beginning to the end of the program.
• Placement. They do not block important visual content such as faces, on-screen text and graphics, do not overlap one another, and do not run off the edge of the screen.

Here is what that looks like in a WebVTT file. The voice tag names the speaker, and the bracketed line carries a sound:

WEBVTT

00:00:01.000 --> 00:00:03.500
<v Maria>Is anyone home?

00:00:04.000 --> 00:00:05.200
[door creaks]

00:00:05.500 --> 00:00:09.000
<v Leo>We should not be here.
Let us go back to the car.

Timing decides whether a caption can be read. For the cues above:

The third cue packs the most text into its time. If a reader falls behind, split it into two cues or give it more time, without letting it drift away from the speech it belongs to.

The rules for television

Under 47 CFR 79.1, distributors and programmers must caption 100 percent of new, nonexempt English and Spanish programming on each channel every calendar quarter, and 75 percent of pre-rule programming. New means first shown on or after January 1, 1998 for analog and July 1, 2002 for digital programming. There are exemptions, among them channels or streams with annual gross revenues under $3,000,000 in the previous year, and interstitials, promos and public service announcements of 10 minutes or less.

The same section treats live and prerecorded programs differently. Small, de minimis errors do not break the standards, and for live and near-live programs the FCC judges accuracy and timing case by case, for example by how understandable the program stayed. For prerecorded programs, the best practices for caption vendors in the rule ask for offline captioning that is a complete textual representation of the audio, including speaker identification and non-speech information, and a style manual so captions look the same throughout.

A viewer who finds missing or broken captions can complain to the FCC or to the distributor, such as the cable or satellite provider, within 60 days after the problem.

The rules for online video

47 CFR 79.4 extends captioning to video delivered over the internet, but only for programming that was shown on U.S. television with captions. Full-length programs have had to keep their captions online since deadlines in 2012 and 2013. Clips followed:

Clips of live programming must be captioned within 12 hours after the program aired, near-live within 8 hours. The rule's definition of video programming excludes consumer-generated media, so a creator's own uploads are not covered by these FCC rules.

That does not make captions optional on the web. WCAG requires captions for prerecorded video with needed audio at Level A and for live video at Level AA. The WCAG basics guide explains the levels.

Making captions and subtitles

1. Get a transcript. For a short clip you can speak the lines into the speech to text tool and correct the result. Add speaker names and sounds by hand, since dictation will not add them.
2. Time it. Split the text into cues of one or two short lines and set start and end times against the audio.
3. Pick the format. The W3C calls WebVTT the most common caption format on the web; SRT and TTML are also used. The subtitle converter turns SRT, WebVTT, SBV and ASS files into each other.
4. Translate for subtitles. Translate the caption text, drop the sound descriptions if the subtitles are for hearing viewers, and keep the timing.

Going the other way, the subtitles to text tool strips numbers and timestamps from a caption file and leaves a clean transcript, which is useful for show notes, search and quoting.

Frequently asked questions

Sources

• W3C WAI: Captions/Subtitles (captions are a text version of speech and non-speech audio; captions same language, subtitles translated; open vs closed; WCAG Level A for prerecorded and AA for live; WebVTT, SRT, TTML)
• 47 CFR § 79.1, Closed captioning of televised video programming, via Cornell LII (definition; 100% of new and 75% of pre-rule nonexempt programming; quality standards for accuracy, synchronicity, completeness and placement; $3,000,000 revenue exemption; interstitials of 10 minutes or less; complaints within 60 days)
• 47 CFR § 79.4, Closed captioning of video programming delivered using Internet protocol, via Cornell LII (full-length programming shown on TV with captions; video clip deadlines; live clips within 12 hours, near-live within 8; consumer-generated media excluded)

Written by Tobias. Updated October 30, 2026.