Captions vs subtitles: what's the difference?

The W3C, which writes the web's accessibility guidelines, draws the line by who's reading:

WCAG defines captions as a text alternative for “both speech and non-speech audio information needed to understand the media content”, and its technique pages add that subtitles give “only the dialogue, in a different human language, and do not include important sounds.” So a line reading [door slams] belongs in captions, and a French translation of the dialogue is a subtitle. Whether either is shown always or switched on by the viewer is a separate question: burned-in vs closed captions.

The split above is the American one: in the US, captions usually means same-language text, and subtitles a translation. In the UK, it's subtitles either way. The BBC's Subtitle Guidelines call the text it makes mainly for viewers with hearing loss subtitles, and say its closed subtitles are “also known as closed captions”. WCAG notes that in some countries captions are called subtitles, and the W3C calls same-language ones intralingual subtitles.

The apps mix the words as well:

• YouTube puts “subtitles and captions” together in its help, and its caption settings say “captions (subtitles)”.
• TikTok said when it launched auto captions in 2021 that they “automatically generate subtitles”.
• Instagram calls the text it writes from a reel's speech closed captions.

So on short-video apps, both words usually mean the same thing: the speech, on screen, in the language it was spoken in.

Captions. A clip of a podcast or a stream is watched in the language it was spoken in, often on mute, so what it needs is that speech on screen, timed to the voice, even when an app's menu calls it subtitles. For clips that means:

• Burned in, so they show on every app and for every viewer.
• A few words at a time, with the spoken word lit up (karaoke captions), rather than whole sentences.
• A sound cue where a sound carries the moment. YouTube's help suggests text like [applause]; the BBC's guidelines say to label only the sounds that matter to understanding, and to describe sounds, not actions.
• Checked. The W3C says automatic captions aren't enough unless they're confirmed to be fully accurate. Read each clip's once with the sound on.

Subtitles are for reaching viewers in another language, usually as a separate track on a full YouTube video, where YouTube Studio lets you add a language. How subtitle files work is in what is an SRT file, and every way to add them in how to add subtitles to a video.