What is a caption?

Two meanings of caption

The word covers two different things, and search results mix them. Before you read on, check which one you need.

Everything below is about the first row. The second is just copywriting: a line or two that sits under the video and does not move with the audio.

What a video caption contains

The W3C Web Accessibility Initiative defines captions as a text version of the speech and the non-speech audio needed to understand the content, synchronized with the audio. WCAG adds that captions go beyond dialogue-only subtitles: they also carry sound effects, music, laughter and who is speaking.

The same W3C page says people use the words caption and subtitle for the same thing in some regions. Its own convention is captions for same-language audio and subtitles for translated speech. For the longer comparison, see captions vs subtitles.

• Speech, in the words actually said.
• Relevant sounds in brackets, such as [applause].
• A speaker label when it is not obvious who is talking.

Caption style and size

Caption style is the look of the text: font, colour, highlight, position and motion. Caption size is how big the text is and how much of it shows at once. The W3C transcribing guide gives one number, and calls it common practice rather than a rule: one or two lines, generally under 32 characters per line.

No standards page I read gives a font size or a pixel safe zone for vertical video. So judge on a phone, at arm's length, with the sound off. WAI also notes that most web players fall back to white characters in a black box, because styling support in players is inconsistent. That is one reason designed captions are usually part of the video itself.

• Keep each line short enough to read in a glance.
• Do not cover faces or on-screen text. WCAG says captions should not obscure relevant information.
• Do not use all capitals for emphasis; W3C advises it only for yelling.

Example: one spoken line, two captions

Say a guest answers: "So we just, um, shipped it on a Friday and, uh, hoped for the best." Both versions below are acceptable. They serve different readers.

Common mistakes

• Trusting automatic captions unread. YouTube says its automatic captions come from machine learning, quality varies, and they should always be reviewed and edited. W3C gives the example that a missing "not" can flip the meaning.
• Putting the whole story in the post caption and leaving the video silent.
• Lines so long that the viewer is still reading when the speaker has moved on.
• Changing or softening what was said. W3C advises transcribing accurately and not censoring.

Caption generators and where KeyMo fits

A caption generator turns speech into timed text, usually with a choice of looks. KeyMo gives each clip word-by-word captions in one of 12 styles (Pro Glow, Bold Pop, Clean, Punch, Classic, Spotlight, Pulse, Karaoke, Reader, Duo, Slide, Headline), or no captions at all. If you want to see what a style does to a short clip, read about word-by-word captions or the wider list of caption styles for Reels and Shorts.