The short answer
Captions are for viewers who can't hear the audio. Subtitles are for viewers who can hear it but don't understand the language, or who just want the words on screen. A caption track carries what a hearing viewer gets from the soundtrack: who is talking, what they say, and the sounds that matter, like applause, music or a fire alarm. A subtitle track usually carries only the words.
Don't trust the label on the track. The W3C's Understanding page for WCAG 1.2.2 notes that in some countries, captions are called subtitles. A track named "subtitles" can be real captions, and a track named "captions" can be dialogue only. What counts is what the track contains.
What a caption track contains that a subtitle track doesn't
A caption track gives a viewer who can't hear the same information a hearing viewer gets. In practice that means a few things a dialogue-only track leaves out:
- Speaker names, when it isn't obvious who is talking: an off-screen voice, a wide shot of a board table, a narrator.
- Sound effects in brackets, such as [applause], [whistle blows], [phone rings] or [gavel strikes].
- Music cues, such as [upbeat music] or [band plays school fight song], so the viewer knows music is playing and what it is.
- Tone and manner cues when they change the meaning, such as [whispering], [laughing] or [sarcastically].
- Silence or interruptions when they matter, such as [audio cuts out] during a livestream replay.
DCMP's Captioning Key, a widely used captioning guide, tells captioners to identify speakers, either by placing the caption under the person talking or by using the same cues a hearing viewer would have, and to describe sounds in brackets.
For example, say a board meeting clip ends with a vote. A dialogue-only track reads: "Motion carries." A caption track reads: "[gavel strikes] BOARD PRESIDENT: Motion carries. [applause]" Only the second one tells a deaf viewer who spoke and how the room reacted.
| Track type | Who it's for | What it contains | Works as captions for WCAG 1.2.2? |
|---|---|---|---|
| Closed captions | Viewers who can't hear the audio | Words, speaker names, sound and music cues | Yes, if accurate and in sync |
| Same-language subtitles (dialogue only) | Viewers who can hear the audio | Words only | Usually not, when speakers or sounds carry meaning |
| Translated subtitles | Viewers who speak another language | Words in another language | No. Translation is a separate question |
| SDH | Viewers who can't hear the audio | Words plus speaker and sound information | Yes, if it really includes them |
| Unedited YouTube auto captions | Anyone, as a rough aid | A machine transcript of the speech | Not on their own |
Where SDH fits in
SDH stands for subtitles for the deaf and hard of hearing. It's an industry term you'll see on streaming services and disc releases. An SDH track is meant to do what captions do, with speaker names and sound cues, but it is formatted and positioned like subtitles.
WCAG doesn't use the term, so the name doesn't decide anything. If a track labeled SDH carries the words, the speakers and the sounds that matter, in sync with the audio, it works as captions. If a vendor delivers an "SDH" file that turns out to be dialogue only, it doesn't. Open the file and read it before you accept it.
Why YouTube says Subtitles/CC
YouTube treats the two words as one thing. YouTube Help covers subtitles and captions in a single article and describes both as files that contain the text of what is said in the video. The same article tells creators to add sound cues like [applause]. In YouTube Studio, every track, whether captions or a translation, is managed under one Subtitles menu, and the player's CC button turns on whichever track the viewer picks.
That means YouTube won't tell you which kind of track you have. You have to open it and look.
Auto captions add one more layer of confusion. YouTube says its automatic captions are produced by machine learning, that their quality varies, and that creators should always review them and edit anything that wasn't transcribed properly. An unedited auto track is a transcript of the speech. It has no speaker names, no sound cues, and often wrong names and terms, so it isn't a substitute for checked captions. More on that in Are YouTube auto-captions ADA compliant?
What WCAG 2.1 AA requires
In 2024, the DOJ's ADA Title II rule set WCAG 2.1 Level AA as the standard state and local government web content usually needs to meet, including public school districts. The ADA.gov fact sheet points out that WCAG includes caption requirements for video. HHS's Section 504 rule adopts the same standard for health centers and other HHS-funded organizations (Section 504 video requirements).
For captions, two success criteria apply:
- 1.2.2 Captions (Prerecorded), Level A: recorded video with audio needs captions. This covers uploads, announcements, patient education videos and the replay of any livestream.
- 1.2.4 Captions (Live), Level AA: live audio needs captions while it streams. YouTube's live auto-captions alone usually fall short. See Livestream captions.
WCAG defines captions as including the dialogue and the non-speech information needed to understand the content. That is why a dialogue-only track usually won't meet 1.2.2. If the viewer can't tell who is speaking, or misses the applause, the alarm or the music that sets the scene, they are missing part of the content.
Translated subtitles are a different matter. WCAG 2.1 AA doesn't require translation. Whether you offer Spanish or other language tracks is a question for your language access plan and your community, not your WCAG checklist. If you do, see Spanish captions for school videos.
This post explains what the rules and standards say. It isn't legal advice.
Common mix-ups in school and clinic video
- Dialogue-only uploads. A caption file built from a script or a meeting transcript has the words but no speakers or sounds. It looks finished and isn't.
- Burned-in translation. Spanish text baked into an English video helps Spanish speakers, but there are no English captions, and nobody can turn the text off or resize it. See Closed captions vs open captions.
- Auto captions renamed or labeled as subtitles. Downloading the auto track and re-uploading it under a new name doesn't make it checked captions.
- A translation with no original-language track. The video has a Spanish track and nothing in English, so a deaf viewer who follows English has no captions at all.
- Music and crowd video with no cues. Concerts, games and graduations depend on sound. A track with only the announcer's words leaves out most of the event. See Captioning school performances.
Do translated subtitles still need captions in the original language?
Yes. A translated track serves people who don't understand the spoken language. Captions serve people who can't hear it. A deaf parent or patient who reads English gets nothing from a Spanish track on an English video, and a translated track rarely carries speaker names or sound cues. Caption the original audio first. Add translations after, if your plan calls for them.
A five-point check for your existing tracks
Pick five recent videos: a board meeting, a game or concert, an announcement, and two others. In YouTube Studio, open Subtitles, select each video, and run through these checks.
- 1
Find the track's source
Look at the language rows. A row that says "(automatic)" is YouTube's machine transcript. If that's the only track, the video doesn't have checked captions.
- 2
Check the language
Make sure there is a track in the language actually spoken in the video. A translation alone doesn't count.
- 3
Look for speaker names
Open the track and read the first few minutes. When the speaker changes or is off screen, the track should say who is talking.
- 4
Look for sound cues
Search the text for square brackets. Applause, music, laughter and alarms should appear as cues like [applause] or [music]. No brackets in a video full of sound is a sign of a dialogue-only track.
- 5
Spot-check accuracy and timing
Play a minute with captions on. Check names, school names and medical terms, and see whether the text appears with the speech. Our caption quality standards post covers what to look for.
Any video that fails one of these checks needs new or edited captions. How to add captions to YouTube videos walks through the upload and editing steps.
Do this before your deadline. Under the Title II rule as extended in 2026, the date is April 26, 2027 for public entities of 50,000 people or more and April 26, 2028 for smaller ones and special districts (ADA.gov). Under Section 504, it is May 11, 2027 for organizations with 15 or more employees and May 10, 2028 for those with fewer. Check yours with the deadline checker.
Start with what's in scope. Video embedded on current pages of your website is in scope. YouTube uploads posted before your deadline are probably exempt (unclear), because neither DOJ nor HHS said whether YouTube counts as social media. Either way, any video someone asks for has to be made accessible. The archived content exception guide explains why old video isn't exempt just for being old.
Checking five videos by hand takes an afternoon. Checking a whole channel doesn't fit in a week for a team of one. Scan your channel to see which uploads have only auto captions or no captions at all.
Sources
- W3C WAI (Understanding SC 1.2.2)
- U.S. Department of Justice, ADA.gov
- YouTube Help (Google)
- YouTube Help (Google)
- Described and Captioned Media Program (DCMP)
This is a reading of the rules, not legal advice. Something out of date? Tell us and we'll fix it.