SDH vs Closed Captions vs Subtitles: What's the Difference?

The difference between SDH, closed captions and subtitles is the audience. Subtitles are for viewers who can hear but do not understand the language, so they carry the dialogue, usually translated. Closed captions and SDH (subtitles for the deaf and hard of hearing) are for viewers who cannot hear, so they add the sounds: music, effects and who is speaking.

SDH and closed captions differ mainly in how they are delivered, and forced subtitles are a fourth kind that switches itself on. All four side by side first, then what each one has to contain.

SDH vs closed captions vs subtitles vs forced subtitles

The column that breaks the pattern is the last one. Forced subtitles are written for hearing viewers who understand the main language, and they are the only kind the viewer never has to switch on.

The other three split cleanly along one line. If the track assumes the viewer can hear the soundtrack, it is a subtitle. If it replaces the soundtrack, it is a caption, whether it is labelled CC or SDH.

What has to be in an SDH track?

Everything a hearing viewer gets from the audio that matters to the story. The W3C's accessibility guidelines define captions as a text alternative "for both speech and non-speech audio information needed to understand the media content" and list sound effects, music, laughter, speaker identification and location as part of that (WCAG 2.2, Understanding SC 1.2.2).

Netflix's English (USA) timed text style guide turns that into concrete rules for SDH:

1. Sound effects and speaker names go in square brackets, in lowercase except for proper nouns: [door creaks], [Maya].
2. Only label what the picture does not show. A speaker ID or a sound effect is used only when it cannot be identified visually.
3. Describe the sound, not the action. "Describe the sounds and audio as opposed to visual elements or actions" — the voice, the speed of speech, the volume.
4. Plot-pertinent sound effects are always included, unless the picture already makes them obvious.
5. Song lyrics sit between music notes, with a space on each side: ♪ like this ♪.
6. Unnamed speakers get generic labels such as [man] or [woman], numbered when there are several.

In a subtitle file, three SDH cues look like this:

12
00:01:04,200 --> 00:01:06,900
[door creaks open]

13
00:01:07,100 --> 00:01:09,800
[Maya] Is anyone here?

14
00:01:10,400 --> 00:01:13,600
♪ I've been waiting all night ♪

Nothing in that file is technically special. It is an ordinary .srt, the same format a translated subtitle uses. What makes it SDH is entirely the content: the bracketed sound, the off-screen speaker named, the lyrics marked as sung.

The same guide gives SDH and translated subtitles the same layout limits: 42 characters per line, and up to 20 characters per second for adult programmes and 17 for children's programmes. Other languages set their own numbers, and some allow SDH to run faster than ordinary subtitles; the reading-speed guide has the table.

In the United States, television captions also carry legal quality standards. The FCC's rules ask for captions that are accurate, synchronous, complete and properly placed, and count background noises and other sounds as part of accuracy (FCC consumer guide).

Is SDH the same as closed captions?

In what they say, almost. In how they reach the screen, no, and that difference is why both names survive.

Closed captions began as a broadcast technology. Apple's Final Cut Pro caption guide describes CEA-608 as "a closed caption standard for broadcasts and web video", and it can travel embedded inside the media file. The player or the TV decodes that caption data and draws it with its own caption renderer.

SDH is the same kind of content delivered as an ordinary subtitle track: an .srt, a .vtt, a track in an .mkv, the subtitle stream on a disc. It can use the fonts and positioning any subtitle can, and it plays on anything that plays subtitles. That is why the same kind of track appears as "CC" on one streaming service and "SDH" on another.

Two more terms complete the set, both from the same W3C page:

• Closed captions "can be turned on and off with some players".
• Open captions "cannot be turned off" — they are part of the picture. Burned-in captions on a TikTok or Reels video are open captions; hardsub vs softsub covers when to burn them in.

The W3C also notes that "in some countries, captions are called subtitles". British broadcasters use subtitles for what American ones call captions, which is the source of most of the confusion in search results.

What SDH is called in other languages

The terms do not translate one to one, so here is what each label means where you will meet it:

Streaming services usually keep the English abbreviation in their menus whatever the interface language, which is why "SDH" and "CC" are searched in every language on this list.

What are forced subtitles?

Forced subtitles are the ones that appear when subtitles are switched off. They cover only what a viewer who understands the main language still could not follow: a scene in another language, a sign, a text message on a phone screen.

Netflix calls them forced narratives, and its English style guide is strict about them: on-screen text is subtitled only if it matters to the plot, in capital letters, never combined with dialogue in the same subtitle, and foreign dialogue is translated only if the audience was meant to understand it.

Inside a video file, what makes a track forced is a flag. The Matroska specification defines the forced flag for subtitles as making a track eligible for automatic selection when it matches the viewer's language "even if the user's preferences would not normally enable subtitles", for "tracks containing only translations of audio in foreign languages or on-screen text". The same specification has a separate hearing-impaired flag for SDH tracks. In MKVToolNix the two are --forced-display-flag and --hearing-impaired-flag; adding subtitles to MKV walks through the muxing, and why a full track must never be flagged forced.

Which one should you make?

"I'm uploading a talking-head video to YouTube." One SDH-style track in the spoken language. Most of it is dialogue, plus the few sounds that matter: [laughs], [upbeat music], an off-screen voice.

"I'm translating my video for other markets." Plain subtitles, one per language. A viewer who reads the Spanish track can usually hear the soundtrack.

"My film has one scene in another language." A forced track for that scene only, plus full subtitles or SDH for viewers who want them.

"I'm delivering to a streaming service or a broadcaster." Their specification decides, down to the bracket style. Netflix publishes a timed text style guide per language.

"It's for TikTok or Reels." Open captions, burned into the picture. Most of the audience watches on mute, so a short sound cue earns its place there too.

Does automatic transcription produce SDH?

No, and it is worth being precise about what is missing. Speech recognition models such as Whisper are built to transcribe speech. They produce the dialogue, which is the largest part of an SDH track, and leave out most of what makes it SDH.

• Sound effects. Whisper does not reliably describe a door, a phone or a gunshot. When it does print something in brackets over music or silence, treat it as a guess to check, not a cue to keep; why Whisper invents text explains where those lines come from.
• Speaker names. Whisper does not know who is speaking. Diarization tools can split voices into "speaker 1" and "speaker 2", but naming them, and deciding when the picture already shows who is talking, is a judgement call.
• Music. Lyrics may be transcribed, sometimes badly; whether a song deserves a ♪ cue, a title in quotes or nothing at all is an editorial decision.

So an automatic transcript is a dialogue subtitle. Turning it into SDH takes one pass by someone who watches the video with the sound on and adds the cues. That pass is quicker than transcribing from scratch, but it is not optional.

Making SDH and subtitles from one transcript

Sablate covers the part automation does well and leaves room for the part it does not. It transcribes the dialogue on your own machine, and the editor is where you add [door slams] or a speaker name to any cue, split a cue that got too long, and drag its timing on the timeline. The editor warns when a line is too fast to read, which matters more in SDH, where sound cues add characters without adding time.

From the same transcript, translations are added as layers, each with its own style and font. That covers both sides of this article: an SDH track in the original language, and plain subtitles in every language you translate into. Each exports as an .srt or .vtt, a burned-in .mp4 for open captions, or, on Pro, one .mkv with a track per language.

Sablate does not identify speakers and does not set the forced or hearing-impaired flags in the .mkv. For the flags, MKVToolNix is the right tool, and for heavy SDH work in dozens of subtitle formats, Subtitle Edit is free:

The free plan covers videos up to ten minutes, three jobs a month and one offline translation layer; Pro is $29 once, and the pricing page lists the difference.

The short version

Subtitles assume the viewer can hear; closed captions and SDH assume they cannot, and add the sounds and the speakers. CC and SDH carry the same kind of content and differ in delivery: broadcast caption data versus an ordinary subtitle track. Forced subtitles are the partial track that shows itself, for foreign lines and on-screen text. If you make one track for your own video, make it SDH in the spoken language, and keep plain subtitles for the translations.

Frequently asked questions

What does SDH mean on subtitles?

SDH means subtitles for the deaf and hard of hearing. An SDH track carries the dialogue plus the sounds a viewer cannot hear: sound effects in square brackets, song lyrics between music notes, and the speaker's name when the speaker is off screen. Streaming menus list it as its own option, next to the plain subtitles in the same language.

What is the difference between closed captions and subtitles?

The audience. Subtitles assume the viewer can hear and usually translate the dialogue into another language. Closed captions assume the viewer cannot hear, stay in the spoken language, and add sound effects, music and speaker identification. Both can be switched on and off; captions that cannot be switched off are called open captions.

Is SDH the same as closed captions?

Nearly, in content; not in delivery. Both describe sounds and identify speakers for deaf and hard-of-hearing viewers. Closed captions in the strict sense are a broadcast format, such as CEA-608 caption data carried in the video signal. SDH travels as an ordinary subtitle track, so it plays wherever subtitles play and uses subtitle fonts and positioning.

What are forced subtitles?

Forced subtitles appear even when subtitles are switched off. They translate only what the audience is meant to understand but cannot, such as a line of foreign dialogue or a sign on screen. Netflix calls them forced narratives. In an MKV file, a forced flag tells the player to show that track automatically when it matches the audio language.

Should I make SDH or regular subtitles for my video?

For a track in the language spoken, make it SDH. Deaf viewers need the sound cues, viewers watching on mute benefit from them, and WCAG defines captions as covering non-speech audio as well as dialogue. Plain subtitles are the right choice for translations aimed at hearing viewers who do not understand the original language.

Can AI generate SDH subtitles automatically?

Only the dialogue part. Speech recognition such as Whisper transcribes what people say, which is most of an SDH track, but it does not reliably describe sound effects, mark music or name speakers. Those cues still need a person who watches the video and adds bracketed descriptions such as door slams and the names of off-screen speakers.