Captions for video ads: 6 rules that work on mute

Captions for video ads work on mute when they follow six rules. Write them before the voice. Keep lines to 42 characters, two lines at most. Break lines where the sense breaks. Give each caption time to be read, about 20 characters a second. Put them on a dark box, clear of the app's buttons. And caption the sounds the story needs, not only the words.

This guide explains where each rule comes from, how to check it with simple arithmetic, and the mistakes that make captions fail. It is for anyone who makes their own marketing video: a launch film, a product explainer, a Reel or a paid ad.

Why captions decide whether a muted video works

In a feed, a video often starts with the sound off. The viewer sees moving pictures and text, and decides in a second or two whether to stay. If the message lives only in the voice-over, a muted viewer gets pictures with no story. Captions are the part of the film that still talks when the sound does not.

The platforms say this in their own guides. TikTok's creative best practices for ads tell advertisers to use captions or text overlays to provide context to your ad. The same page asks you to introduce your content proposition in the first 3 seconds. On mute, the first caption is the only place where that can happen. It is part of the hook, as our post on openings that stop the scroll explains.

Captions are also access. The W3C notes that captions provide content to people who are Deaf and hard-of-hearing, and that they are also used by people who process written information better than audio. A caption that is good for a muted scroller is good for these viewers too. So the work pays twice.

Where the rules come from

Ads have no caption standard of their own. The best rules come from the people who have subtitled film and television for a long time, and from the web's accessibility standard. Two sources do most of the work in this guide.

The first is Netflix's timed-text style guide for English. It sets limits for line length, line count, line breaks and reading speed, and a separate page sets timing rules. These limits were made for long programmes, where a viewer reads for an hour. An ad is short, but the eye reads at the same speed, so the limits hold.

The second is the W3C, which writes WCAG, the web's accessibility standard. It defines what a caption must contain. It also explains the difference between captions that a viewer turns on and captions that are part of the picture. For a feed ad, that difference decides how you deliver the file.

Platform guides add one number of their own: TikTok's words per second. Where the sources give two limits for the same thing, this guide takes the stricter one.

Rule 1: write the captions before the voice

Most captions fail because they are a transcript added at the end. The voice-over was written for the ear, with long sentences and filler, and the captions copy it word for word. On mute, that is a wall of text.

Turn the order around. Write the lines on screen first, as if the film had no sound at all. Then write the voice to match them, word for word or close to it. Two FoxyLab Showcases go one step further and have no narrator. The Ritual Coffee launch ad says everything in two short lines and a button. In the Crumb & Co croissant ad, every line is set on the picture, so it reads on mute.

A quick test: read only the captions, top to bottom, with the picture hidden. If they tell the story and end on the action you want, the film works on mute. If they read like pieces of a conversation, rewrite them.

Rule 2: keep each line to 42 characters, two lines at most

Long lines make the eye travel, and on a phone they run into the edge of the frame. Netflix's English style guide sets the limit at 42 characters per line, with a maximum two lines. Use that as a ceiling. On a 9:16 screen, where the app covers part of the frame, you will often want less: our guide to the Reels ad safe zone shows how much space stays clear.

Count characters, not words. Spaces and punctuation count. "You still have no plan." is 23 characters, so it fits on one line with space to spare. If a caption needs a third line, the sentence is too long. Split it into two captions, or cut words until it fits.

Short lines have a second benefit. A short line can be set larger, so it reads at the size of a phone held at arm's length. And it leaves room for the product, which is what the film is there to show.

Rule 3: break the line where the sense breaks

When a caption needs two lines, the place of the break decides how fast it reads. Netflix says to break lines after punctuation marks, before conjunctions and before prepositions. It also lists pairs never to split, such as a noun from an article and a verb from a subject pronoun.

• Good: "Tap what is in / your fridge." Bad: "Tap what is in your / fridge."
• Good: "Plan dinner, / not stress." Bad: "Plan dinner, not / stress."
• When both breaks are correct, make the bottom line the longer one. Netflix asks you to prefer a bottom-heavy pyramid shape, but avoid having just one or two words on the top line.

Automatic captions get this wrong most often. They break where the line runs out of room, not where the meaning ends. Check every two-line caption by hand.

Rule 4: give each caption time to be read

A caption that leaves before it is read is worse than none: the viewer gets half a sentence and loses the thread. Two published numbers set the speed. Netflix allows up to 20 characters per second for adult programmes and up to 17 characters per second for children's programmes. TikTok says: we recommend displaying 5-10 words per second when using text. Words differ in length, so characters give the safer count.

Netflix's timing guide adds a floor: subtitles should not be any shorter in duration than 20 frames (or 4/5 sec). It also asks for a minimum of 2 frames between them, so that the eye sees a new caption arrive.

A worked example: six captions in a 15-second film

Platewise is a fictional meal-plan app. Its 15-second film has six captions. Divide each caption's characters by 20 to get its minimum time on screen, and raise any result under 0.8 seconds to the floor.

Caption	Characters	Minimum time
Dinner is at six.	17	17 ÷ 20 = 0.85 s
You still have no plan.	23	23 ÷ 20 = 1.15 s
Open Platewise.	15	15 ÷ 20 = 0.75 s, raised to 0.8 s
Tap what is in your fridge.	27	27 ÷ 20 = 1.35 s
Get three meals in ten seconds.	31	31 ÷ 20 = 1.55 s
Platewise. Plan dinner, not stress.	35	35 ÷ 20 = 1.75 s
Total	148	7.45 s

The sum is 0.85 + 1.15 + 0.8 + 1.35 + 1.55 + 1.75 = 7.45 seconds. The film runs 15 seconds, so 15 − 7.45 = 7.55 seconds are left. That time lets each caption stay longer than its minimum and lets the picture work alone between lines. At the children's limit of 17 a second, the same 148 characters need 148 ÷ 17 ≈ 8.7 seconds, which still fits.

If the total comes out longer than the film, the script has too many words. Cut words. Do not make the captions faster.

Rule 5: make captions readable on any picture

White text with no background disappears on a white kitchen counter, a bright sky or a light product shot. Put the caption on a solid dark box, or give it a strong outline. Use a bold sans-serif type. Keep the size and the place the same through the whole film, so the eye knows where to look.

The place matters as much as the style. In a vertical feed, the app lays its own text, buttons and icons over the film. A caption in the lower third, where film subtitles usually go, can end up under them. Put captions in the middle band of the frame, inside the safe zone.

There is also a reason to burn the captions into the picture. The W3C explains that most captions on the web are “closed captions” that can be hidden or shown by people watching the video, while open captions are always displayed and cannot be turned off. About styling closed captions, it says: support in browsers and other media players is inconsistent and sometimes unreliable. Captions in the picture look the same for every viewer in every feed. Where a platform also takes a caption file, upload one, for viewers who rely on their own settings.

Rule 6: caption the sounds, and check every automatic caption

Captions are not only the words. The W3C defines them as a text version of the speech and non-speech audio information needed to understand the content. WCAG makes captions for recorded video a Level A requirement, and says captions identify who is speaking and include non-speech information conveyed through sound. If the joke of the film is a doorbell, write "[doorbell]" on screen. If two people talk, make it clear who says what.

Automatic captions save time, but they are a first draft. YouTube says it uses speech recognition technology to automatically create captions for your videos, and that automatic captions will be in the video's default language only. The W3C is direct about them: you will need to edit automatic captions for accuracy. Read every one for the brand name, numbers and prices. A wrong price in a caption is a wrong price in the ad.

Common mistakes and their fixes

• The captions are the transcript. Fix: write the lines on screen first, then cut each to one idea.
• Three or more lines on screen. Fix: split the sentence into two captions, or cut words until it is 42 characters or fewer a line.
• A word alone on the top line. Fix: move the break so the bottom line is the longer one.
• Captions change with every word. Single words that flash on screen look busy and give the eye no time. Fix: show a whole phrase and keep it for at least 0.8 seconds.
• Thin white type on a bright picture. Fix: a dark box or a strong outline, in the same place for the whole film.
• The caption sits under the app's buttons. Fix: move it into the middle band of a 9:16 frame and check it on a phone.
• A brand name spelled the way it sounds. Fix: read every automatic caption before you publish.

A checklist before you publish

1. Watch the film once on mute, on a phone. Does it make sense?
2. Count the characters in the longest line. Is it 42 or fewer?
3. Add up all the characters and divide by 20. Is the result well under the film's length?
4. Find every caption under 0.8 seconds. Make it longer, or join it to the next one.
5. Look at every two-line caption. Does each break fall where the sense breaks?
6. Check that the captions sit clear of the app's buttons, on a box or an outline.
7. Check the brand name, the numbers and every sound the story needs.

Questions about captions for video ads

Should I burn captions into the video or upload a caption file?

For feed video, burn them in. Open captions are always shown and look the same in every player. Upload a caption file as well where the platform takes one, for viewers who use their own caption settings.

How many words can a caption have?

Count characters instead: up to 42 characters a line and two lines at most. On a phone, one line of about 20 to 30 characters is often easier to read.

How long should a caption stay on screen?

Divide its characters by 20 to get the minimum in seconds, and never go below 0.8 seconds. A 27-character caption needs at least 1.35 seconds.

Are automatic captions good enough?

They are a first draft. Edit them for accuracy, especially brand names, numbers and prices, and fix the line breaks by hand.

Make the captions part of the film

Captions work best when they are written with the script and timed with the cut, not typed over a finished edit. FoxyLab works this way. You paste a link or write a line in the chat. The studio writes, voices, animates and renders the film in every size, from 9:16 to 16:9, and changes it from your notes, captions included.

Take your next script, count its caption characters, and divide by 20 before anything is animated. Then open the studio, paste your product's link, and ask for a film whose captions tell the whole story on mute.