Subtitles vs captions: what is the difference?

Three jobs for the same strip of text

The useful way to separate the terms is to ask who the text is for. A caption stands in for the soundtrack, so it has to carry everything a hearing viewer would take from the audio: the words, who says them, and sounds that change the meaning, such as a phone ringing or music turning tense. A subtitle stands in for language knowledge, so it assumes the viewer hears the sounds and only needs the dialogue in a language they read. The W3C puts it simply: captions are for audio in the same language, subtitles for spoken audio translated into another language.

What goes into each

Captions include dialogue, speaker labels where the speaker is off screen or unclear, and short sound descriptions in brackets, like applause or a door closing. Music is noted when it matters, for example a song that carries a lyric the story depends on. Subtitles usually contain dialogue only, and translators often condense lines to fit a comfortable reading pace, because translated sentences can run longer than the original. SDH follows subtitle conventions for timing and layout but adds the speaker labels and sound descriptions captions carry, so that a track made for a film release also serves viewers who cannot hear it.

The words change between countries

Much of the confusion is regional. In North America, captions means same language text for accessibility, and subtitles means translation. In the UK and much of Europe, subtitles is used for both, and the BBC publishes its accessibility guidance as subtitle guidelines, covering timing, colour and sound labels for deaf and hard of hearing viewers. SDH grew up as a label on streaming services and disc releases, where a subtitle file needed to signal that it included sound information. When a brief, a contract or a platform setting uses one of these words, check which meaning is intended before delivering a file.

When it matters

It matters whenever a video carries information someone must receive. The W3C accessibility guideline for prerecorded video asks for captions on synchronised audio, which a translated subtitle track does not satisfy on its own, because it leaves out the sounds. It matters for reach too: a video sold into another market needs subtitles or a new voiceover, and captions in the original language do nothing for that audience. And it matters in feeds, where many hearing viewers watch with the sound off and read captions simply because audio is not an option in that moment.

Common mistakes

Labelling a translated subtitle file as captions is the most common, and it leaves deaf viewers without sound cues. Publishing machine transcribed captions without a check is the next, since product names, numbers and accents are exactly where recognition goes wrong. Summarising instead of transcribing turns captions into a paraphrase that no longer matches the lips on screen. Mixing languages within one track, such as translating some lines and leaving others, confuses both audiences. And leaving out a crucial sound, such as a warning tone in a safety video, removes the one piece of information that viewer most needed.

Where it shows up in the product

The video builder here makes captions in the strict sense: they are generated from each scene's narration text, so they are always in the language that scene is narrated in, and they carry speech only, with no sound descriptions or speaker labels. Language can be set per scene, so captions follow if narration switches language partway. There is no separate subtitle track and no way to caption in a language different from the voice. To reach another market, build a version narrated in that language, as the Spanish restaurant example does, and the captions arrive in Spanish with it. The captions are burned into the picture rather than supplied as a file.

Questions people ask