How to Decide Which Sounds Deserve a Caption Description

When I teach sound descriptions, I begin with a single test. If a viewer who cannot hear this sound would misunderstand the scene, miss a reaction, or lose a thread of the story, it earns a description. If the picture and the speech already carry the meaning, it does not. That test removes most of the guesswork, and it keeps captions from filling up with noise nobody asked about.

Consider a recorded interview in a quiet office. A phone rings and the speaker pauses, apologises and silences it. The ring explains the pause, so a short note such as [phone rings] helps. The soft hum of an air conditioner, on the other hand, changes nothing about what is said, so it stays out.

Work through your first pass of a video with this test in mind. Mark each moment where a sound seems important, then come back and decide one by one rather than writing descriptions as you listen.

• Ask what the viewer would miss without the sound
• Keep sounds that explain a pause, a reaction or an action
• Skip steady background noise that changes nothing
• Decide after a full pass, not while listening

In recorded talks and event footage, a handful of sound types come up again and again. Audience reactions are the most common: laughter, applause and murmurs all tell the viewer how a remark landed. A speaker who says something dry and then waits makes little sense on screen if the laughter that follows is missing from the captions.

Interruptions matter too. A door closing, a microphone thumping, a phone ringing or a sudden loud crash can change what the speaker does next. Music that begins or ends a segment is worth noting when it sets a mood or marks a transition, especially at the start and end of an event video.

Off-screen sounds deserve extra care. If someone speaks or reacts from outside the frame, the viewer has no visual clue at all, so the caption has to supply it. Keep a short list of these sound types beside you while you work, and check each against the test from the previous section.

• Laughter, applause and audience murmurs
• Interruptions: phones, doors, microphone noise
• Music that opens, closes or shifts a segment
• Off-screen sounds the picture does not show

A good description is short, plain and placed in square brackets so it reads as a note rather than speech. Aim for the fewest words that still name the sound: [applause], [laughter], [door slams]. Use a verb or a noun the viewer recognises at once, and avoid stacking adjectives that try to paint the sound for them.

Add detail only when detail changes the meaning. [Soft laughter] and [Loud laughter] mean different things in a panel discussion, while [sighs] and [sighs heavily] may not. If a sound is repeated, describe it the first time and trust the viewer afterward, rather than marking every single instance.

Write in the present tense and keep the wording neutral. Describe what is heard, not what you think the speaker feels. [Voice breaks] is a sound; [sadly] is an opinion. Your job is to report, and the viewer draws the conclusion from the scene.

• Use square brackets and lower-case plain words
• Name the sound, not the emotion behind it
• Add detail only when it changes the meaning
• Describe a repeated sound once, then let it go

A sound description is still a caption, so it needs a start time, an end time and a readable length. Place it so it appears when the sound happens, not a moment before or after. A late [applause] feels strange when the clapping has already faded, and an early one gives away the moment.

Give each description enough time on screen to be read, usually the same minimum as a short line of speech. If a sound lasts a long while, such as sustained music, a single note at the start is often enough, and you can mark its end if the change matters.

When a sound overlaps with speech, keep both readable. Put the description on its own line or in its own caption block rather than squeezing it into a sentence. Preview the result at normal speed, with the sound on and then off, to confirm that the timing feels natural in both cases.

• Match the start time to the sound itself
• Allow enough time on screen to read it
• Mark long sounds once, and their end if it matters
• Keep descriptions apart from spoken lines

Once the descriptions are in, watch the video with the sound off and the captions on. This is the closest you can get to the viewer's experience. Ask whether you always know what is going on, and whether any bracketed note interrupts your reading without adding anything.

Then watch again with the sound on and look for missed moments. Often a pause in the speech has no explanation in the captions, and a short note fills the gap. Equally, you may find descriptions that repeat what the picture already shows, and those can go.

Keep a consistent style across the whole file. If you wrote [applause] early on, do not switch to [audience claps] later. Consistency makes the captions feel calm and professional, and it makes the later quality review much faster for whoever checks them.

• Watch once with sound off, once with sound on
• Remove notes that repeat the picture
• Fill unexplained pauses with a short note
• Use the same wording for the same sound

Pick a short recorded talk or interview, perhaps three to five minutes long, and play it through once without writing anything. Listen only for sounds that are not speech and jot them on a sheet of paper in the order they appear, without judging them yet.

Now apply the viewer test to each item. Cross out the ones that change nothing, and write a bracketed description of five words or fewer for the ones that remain. Set start and end times in your caption tool and play the result with the sound off.

Finally, hand it to someone who has not seen the video and ask what they understood. Their questions show you exactly where a description is missing or unclear, and a few rounds of this teach more than any rule list could.