Captions for deaf and hard-of-hearing audiences
By AutoCap · 31 October 2026
What deaf and hard-of-hearing viewers need beyond a transcript: speaker names, sound cues, faithful wording, timing and contrast, and what AutoCap can add.
More than a record of the speech
Captions on short videos are often written for sound-off scrolling. For a viewer who is deaf or hard of hearing, they are the soundtrack, and a transcript of the speech misses much of what hearing viewers take in unnoticed: who is talking, a laugh that marks a joke, music that turns a scene tense, a knock everyone on screen reacts to.
Good captions for these viewers carry everything in the audio that affects the meaning, faithfully. That takes no special software, only a careful pass through the video.
What to add to the words
Add these where the video needs them. I'd write cues in the same language and script as the captions, so a Devanagari transcript doesn't switch to English every time a door closes.
- Who is speaking, when the picture doesn't show it: a voice off camera, a narrator, two people in one shot. A name at the start does the job, as in 'Riya: chalo, shuru karte hain', and a new speaker should start a new caption.
- Sounds and music that carry meaning, in square brackets: [door slams], [laughter], [soft piano], [music stops], plus the lyrics if the video depends on them. Skip sounds that change nothing, or the cues crowd out the dialogue.
- How something is said, when the words alone would mislead: [sarcastic] or [whispering] before the line.
- The words actually spoken. Tidying a stammer is a judgement call; summarising a sentence, dropping a joke's setup or softening the speaker's word choice changes what the viewer is told.
Timing, motion and contrast
A viewer who relies on captions reads every one, so each must stay up long enough to finish; 'Min duration' under Caption settings holds short ones longer. Word-at-a-time styles are hard going because the sentence never appears whole, so where meaning matters I'd choose a phrase or stacked template.
Captions should arrive with the speech; why captions end up out of sync covers how to check that. Text is also harder to read while it is still moving, so consider setting Entrance to None in the Motion pane.
Pale text needs a dark edge or a solid block behind it to stay readable over bright or busy footage. Where a template allows them, the Effects pane has switches for 'Drop shadow', 'Text stroke' and 'Background'. Keep captions off the speaker's mouth, since some viewers lip-read alongside the text, and don't use colour as the only sign of anything, because not every viewer can tell colours apart.
Burned-in or closed captions
Burned-in, or open, captions are part of the picture: every viewer sees them, on any app or device, styled as you designed them. Nobody can switch them off or resize them, a screen reader can't read them, and correcting one means exporting and uploading again.
Closed captions are a separate text track the player shows when the viewer asks. They come from a caption file, and AutoCap's paid plans download two kinds: SRT, which YouTube and editors such as Premiere and Resolve accept, and VTT, used by video players on web pages. Viewers can turn them on and off, some players let them pick size and colour, and neither file carries your fonts, colours, position or animation.
If you use both, a viewer who switches on closed captions over burned-in ones sees two sets of text. The files also follow the on-screen grouping: downloaded under a One word template, every cue holds a single word, so switch to a Single line template first.
What AutoCap does, and what you add
AutoCap transcribes speech. It doesn't label speakers or describe sounds, so those cues are yours to add, and anything you add goes into the SRT and VTT files too. There are two ways:
- For a sound in a gap, press '+ Line' at the top of the Captions panel to add a caption you type yourself, at the playhead or in the next pause with room for one. It is marked TYPED, has In and Out times you can set, and stays put if you re-sync to the audio.
- For a speaker's name, or a sound under speech, double-click the caption's first word (or select it and choose 'Edit text') and type the cue in front of it. The word keeps its timing, so the cue appears with it.
Where this approach falls short
AutoCap can't tell that a slammed door matters to the story or that a voice belongs to someone off camera, and a misheard word that is also a real word looks normal in the caption list, so read the captions against the audio from start to finish.
Typing a cue into a word is a workaround, not a feature built for cues. Only one caption shows at a time, so a typed caption can't overlap speech, and a cue typed into a word behaves like the word: on One word templates it shows only while that word is up, and styles that highlight the spoken word light it up too.
This is widely used accessibility practice, not a legal or broadcast standard, and some platforms and clients ask for more. The best test is a viewer who relies on captions telling you where they got lost; failing that, watch with the sound off and note every moment you had to guess.
Questions
Are automatic captions enough for deaf viewers?
Not on their own: they cover speech, not speakers or meaningful sounds, and they can mishear words.
How do I show a sound effect in captions?
Describe it in square brackets, like [door slams], when it happens. In AutoCap, use '+ Line' if it falls in a pause, or type it in front of a word if it happens during speech.
Are burned-in captions the same as closed captions?
No. Burned-in captions are drawn into the video and stay on for everyone; closed captions come from a separate file, such as SRT or VTT, that the viewer can turn on or off.
Should captions be in the spoken language or translated into English?
For viewers who read the spoken language, native script or romanised text keeps the speaker's own words. AutoCap's English translation gives the meaning in English instead, for viewers who don't know the language.