What 'Non-Speech Information' Means in Captions, in Plain Words

In captioning, the phrase non-speech information sounds heavier than it is. It simply means every sound a viewer would normally hear that is not someone talking: applause, a phone ringing, a door closing, music starting, a chair scraping across the floor. When we write these sounds into a caption file as short bracketed notes, we are giving viewers who cannot hear the audio the same clues everyone else gets for free.

The idea is old in the trade, and you will also see it called sound description or audio description of sounds. The wording changes between style guides, but the job stays the same. A caption track that carries only dialogue tells part of the story. One that carries dialogue plus the sounds that shape it tells the whole story, and the second version is the one a person who is deaf or hard of hearing can actually follow.

I teach this inside the Sound Descriptions module, and beginners usually relax once they see it is a judgment about usefulness, not a duty to write down every noise.

The working rule I give learners is a single question. If a viewer could not hear this sound, would they miss something that affects how they understand the scene? If the answer is yes, describe it. If the answer is no, leave it out. A room full of people laughing after a speaker's joke passes the test, because without it the pause that follows looks strange.

A refrigerator humming in the background of an interview fails the test. It is real, but it adds nothing, and writing it down would only slow the reader. Captions are read in a few seconds per line, so every bracketed note competes with the spoken words for the viewer's attention. That is why restraint matters as much as completeness.

When I review a learner's first attempt, the most common correction is deletion. People hear everything because they are listening closely, and they want to prove it. A calm, short track usually reads better than a thorough one.

• Describe sounds that carry plot, mood or information
• Skip steady background noise that never changes
• Ask whether the viewer would feel lost without the note

A good description is short, in the present tense, and names the sound rather than interpreting it. Square brackets are the usual convention: [applause], [phone buzzing], [door slams]. Use plain nouns and verbs. Write what the ear would pick up, not your opinion of it, so [tense music] is acceptable where music clearly signals tension, but [annoying beeping] adds a judgment the viewer did not ask for.

Keep each note to a few words and place it on its own line when it stands apart from speech, or at the start of a caption when it leads into a sentence. If the same sound repeats, such as a ringing phone, describe it fully the first time and shorten it afterward so the reader is not buried in repetition.

Consistency counts here. If you wrote [laughter] in the first ten minutes, do not switch to [chuckling] and [audience laughs] without a reason. A reader learns your vocabulary quickly and trusts it, so changing it silently creates small moments of doubt.

• Use square brackets and lowercase unless your style guide says otherwise
• Name the sound, avoid adjectives that express opinion
• Shorten repeats after the first full description

Music is where beginners hesitate most. If a song is part of the scene, for instance a band playing at an event, note it once when it begins and again when it ends, and add the title only if it is known and relevant. Background music under a talk usually needs no note at all, unless it swells and drowns out the speaker.

Off-screen speech is a different case, because it is not non-speech but still needs a signal. A label such as [man off screen] or the speaker's name tells the viewer who is talking when no one is visible. Whispering, shouting and speech in another language also deserve a short tag when the viewer would otherwise be confused by the change.

Some sounds are ambiguous, and you may not be able to identify them. In that case, describe what you can honestly hear, such as [muffled noise], instead of guessing. A truthful vague note is better than a confident wrong one, since captions are read as fact.

• Mark music when it starts and ends if it matters to the scene
• Tag off-screen voices so viewers know who is speaking
• Prefer an honest vague note over a confident guess

A sound note takes reading time just like dialogue, so it has to fit the timing of the caption it sits in. If a door slams during a pause, the note can occupy that pause comfortably. If it lands in the middle of rapid speech, you may need to merge it into a neighbouring caption or drop it, because a line that appears for under a second cannot be read.

I ask learners to play the video with the sound off and read only the captions. This simple exercise exposes notes that arrive too late, linger too long or crowd the screen. It also shows you whether the track still makes sense without audio, which is the real test of the work.

Place the note close to the moment of the sound. A [crash] that appears two seconds after the crash confuses the viewer more than leaving it out would, because the cause and effect have drifted apart on screen.

Take a three-minute clip of an interview or event recording and watch it once with the sound on, noting every non-speech sound you hear. Then watch again and cross out each sound that does not change meaning. What remains is usually three to five notes, and those are the ones worth writing.

Write each remaining description in five words or fewer, then play the clip with sound off and check that every note appears at the right moment. Ask a colleague who has not seen the video to read the captions and tell you what happened. Their confusion, if any, points to the notes you got wrong or missed.

This loop of listening, cutting, writing and testing is the core habit of the module, and it works on any recording, from a short interview to a long panel.