Regular subtitles assume you can hear the mix and only need the words translated or transcribed. SDH assumes you cannot. It adds labels for who is speaking when that is not obvious, and descriptions of sounds that carry plot or tone: a knock, a siren, a song that is diegetic rather than a bed. Those brackets are the whole difference. The file format is usually the same sidecar as closed captions; the content of the cues is not.
Accessibility standards treat SDH as captions, not as 'nice extra subtitles.' A course, a broadcast master, or a public-interest clip that only carries dialogue has not done the caption job for a deaf viewer. Short-form ads rarely ship full SDH because the format is already burned in and already designed to be read on mute — but a talking-head with off-screen sound that matters still needs the label, or the mute viewer misses the beat the edit is built on.
Do not confuse SDH with forced narrative (only the lines a same-language viewer still needs) or with auto captions (a transcript of speech, no [music] line unless you wrote one). Auto-transcribe will not invent sound descriptions. If the brief is SDH, someone still has to author the non-speech cues.
In practice
- If a sound causes a cut or a joke, write it into the SDH track; a mute viewer will not infer it.
- Identify speakers when two voices share a frame or when the speaker is off screen.
- Auto captions give you the speech. SDH is a pass after that, not a preset.
The mistake to avoid
Shipping a speech-only transcript and labelling it SDH. Without non-speech cues and speaker IDs it is just subtitles.
Where you will run into it
- Add Captions to a Video — Speech in, styled subtitles out — no manual timing.
- AI Caption Generator — Drop a video, get timed captions. Free, on your device, no upload.
Related terms
Closed captions
Closed captions are a separate subtitle track the player draws, which the viewer can switch off and which each platform styles itself.
Burned-in captions
Burned-in captions meaning: subtitles rendered into pixels, so viewers cannot turn them off or lose them on re-upload. Closed captions are a separate track.
Forced narrative
Forced narrative captions only translate what a same-language viewer still cannot read — a sign, a letter, foreign speech — not the whole dialogue.
Speech-to-text
Speech-to-text meaning: converting spoken audio into written text, the transcript captions, translation, and search all depend on.
Hook rate
Hook rate meaning: the share of people shown a video who are still watching a few seconds in. It grades the opening, not the body.
The all-in-one AI studio for creators. 60+ models for video, image, voice, music and lipsync, in your browser or on your phone.
Free on the web and on iPhone, one account.