Subtitles assume the viewer can hear the audio and only need the words: the dialogue, transcribed or translated. Closed captions assume the viewer cannot hear it, so they also carry who is speaking and the sounds that matter — [door slams], [phone buzzing], [tense music]. "Closed" means the viewer can switch them off, because they travel as a separate track or file. Open captions are burned into the picture and are always on. The file formats are the same; what differs is what you write in them and how you deliver them.
Who is the text for — someone who can hear, or someone who cannot? And can the viewer turn it off?
| Written for | Contains | Can be switched off | |
|---|---|---|---|
| Subtitles | Viewers who can hear but need the words, often in another language | Dialogue only | Usually, as a separate track |
| Closed captions | Viewers who cannot hear the audio | Dialogue, speaker labels and meaningful sounds | Yes — that is what "closed" means |
| Open captions | Everyone, whether or not the sound is on | Usually dialogue, sometimes sounds | No — they are part of the picture |
| SDH | Viewers who cannot hear, in a subtitle track | Dialogue, speakers and sounds, styled like subtitles | Yes |
Usage is not consistent everywhere. In British English "subtitles" often covers all of these, and plenty of platforms label any text track "CC" whether or not it describes a single sound. When someone hands you a brief, ask which of the two questions above they care about, because the name on the brief may not tell you.
SDH — subtitles for the deaf and hard of hearing — is the hybrid you meet on streaming services and discs: caption content (speakers, sounds) delivered in the look and the track slot of a subtitle. For most people making their own videos, the useful distinction is simpler: dialogue only, or dialogue plus sound, and a file, or burned in.
This is the choice with practical consequences, and it is made at delivery, not while writing.
Closed captions are a separate file — an SRT or a VTT — or a track inside the video container. The player draws them on top of the picture, the viewer turns them on or off, and the platform can read the text. That last part is easy to overlook: a caption file is text a platform can index and offer in other ways, while a caption burned into pixels is just pixels.
Open captions are rendered into the picture before the video is uploaded. They are always visible, they look identical on every player, and they need no support from the platform at all. They also cannot be turned off, cannot be corrected without re-exporting the video, and are not text to the platform any more.
So the usual split:
<track> element, covered in
converting SRT to VTT.The extra content is small, and it is the part a speech model will not write for you.
A caption has to give someone who cannot hear the audio everything they need to follow it. In practice that is three additions to the dialogue:
[MAYA] Is it on?[door slams], [phone buzzing], [laughter]. The
test is whether a hearing viewer would react to it. The fridge humming in the background
does not need a caption; the knock at the door does.[upbeat music],
[tense music], [music stops]. If the lyrics matter, caption
the lyrics.Keep sound descriptions short and in the present tense, and give them their own line or their own moment rather than squeezing them into a dialogue line that is already at its reading-speed limit. A caption track is only useful if it can be read in the time it is on screen — subtitle reading speed has the numbers, 42 characters a line and about 20 characters a second, and they apply to bracketed sounds as much as to speech.
If you are captioning to meet an accessibility obligation, check the standard that applies to you for its exact rules; this page is about the craft, not the law.
suubbs writes the dialogue; you add the rest; the export decides open or closed.
.ass) from the ··· menu, and burn it into the picture with one ffmpeg
command. suubbs has no one-click burn-in button yet;
burning subtitles into a video has the command
and the four problems everyone hits.The free browser version does all of this on the first 60 seconds of a video. The Mac app does the whole video, with a larger model and word timing measured at about 3 ms.
Short answers to the follow-up questions.
Subtitles carry the dialogue for viewers who can hear the audio. Closed captions are written for viewers who cannot, so they also identify speakers and describe meaningful sounds and music. "Closed" means they are a separate track the viewer can turn off.
Closed captions are a separate file or track that the player draws and the viewer can switch off. Open captions are burned into the picture, always visible and identical on every player, but they cannot be turned off or corrected without re-exporting the video.
Either. SRT is a container for timed text, and what makes it captions is what you write in it — speaker labels and sound descriptions as well as dialogue. Uploaded beside a video it is closed, because the viewer can turn it off.
Subtitles for the deaf and hard of hearing: caption content, with speakers and sounds, delivered in the style and track of a subtitle. It is common on streaming services and discs.
Usually open, burned in, because many people scroll with the sound off and a styled caption is part of the video's look. Where the platform also accepts a caption file, uploading one as well costs nothing.
Do not count on it. A speech model is built to write what is said. In suubbs, speaker labels and sound descriptions are added by hand, as a new line inserted at the playhead, and on a layer of its own on the timeline when it needs to overlap dialogue.