Captions and subtitles: how to make a business film work with the sound off

Captions are usually treated as packaging: something done after the film is finished, by somebody who was not in the room when it was written. That is why so many business films fail the people who cannot hear them.

A film that carries its message in the picture as well as the voice survives being watched silently, on a train, in a meeting room, in a feed.

Captions and subtitles are not the same thing

Captions are for people who cannot hear the film. Subtitles are for people who can hear it and cannot understand the language. That sentence is the whole of the captions vs subtitles argument, and getting it wrong is the most common mistake in business video.

Captions carry the words, and what the words cannot: a speaker's name when two people talk, a sound that matters, and music telling you how to feel. Subtitles translate speech and nothing else.

In practice you need both. Same-language captions for the audience you already have, subtitle files for the languages you want to reach. Ask for them as files rather than a burnt-in layer, so a player can switch them on and off. The common formats are SRT and WebVTT, and the answer to which one is usually both.

Most of your audience is watching with no sound

People watch business films where sound is not available or not polite: on a train, at a desk in an open office, in a meeting where somebody is talking, in a feed where every other video is silent. That applies to more people than hearing loss does.

Which gives one rule worth designing around: if the message only exists in the voiceover, it does not exist for most of the audience. There are two honest answers. Build the text into the film, so it reads without a caption file. Or accept that the film depends on narration, and make the captions good enough that switching them on costs nothing.

The first is better and has to be decided before the film is built. A film designed around a voiceover is hard to caption well afterwards, because captions over an animation composed to be empty at the bottom always look like what they are. Either way the captions have to exist, because leaving them out is the part that fails.

What buyers now ask for, and how to put it in a tender

Accessible media used to be a nice line in a proposal. It is now a line in the requirements, particularly in the public sector and in large organisations, and the pattern is consistent: if the film is published on an organisation's own website or used in a service the public meets, somebody will ask who made it accessible.

Two things are worth knowing without reading any legislation. The recognised technical standard is WCAG, the Web Content Accessibility Guidelines, now at version 2.2, and its video requirements are concrete: prerecorded video with sound needs captions, video that carries information only in the picture needs an audio description or an equivalent alternative, and a full text alternative such as a transcript is expected. And the people running your tender usually do not want a legal argument. They want a deliverable they can point at.

So turn accessibility into procurement language and write it into the brief as deliverables: caption files, subtitle files, a transcript, an audio-described version if the film carries information visually. Ask for the exact formats, ask whether captions are written from the audio or from the script, and ask what happens when the film is revised. A caption file that is a week out of date is worse than none.

Do not accept "we will look at accessibility later". Later is after the composition is locked, which is exactly when it becomes expensive. WCAG video requirements give you a checklist that does not depend on taste, and in a tender that is worth more than any promise about quality.

Designing so captions do not fight the picture

Captions ruin a film in two ways: they cover the thing you are meant to look at, or the picture leaves them nowhere to live and they float over the busiest part of the frame.

Both are composition problems, so both are solved at the storyboard. Decide where text is allowed before anything is drawn. A film with one focal point per shot can hold a caption band at the bottom without a fight. A film with a busy chart, a moving diagram and a logo in the lower third cannot.

The other half of the problem is that one film is rarely one shape. A film is often wanted wide, tall and square, and the caption safe area is different in each. A band that sits comfortably in a wide frame cuts through the middle of a vertical one, and in 9:16 the bottom of the frame belongs to the app, so a caption there ends up under the interface and the platform's own buttons.

So pick the safe area per shape and hold it. In 16:9 at 1920 by 1080, keep captions inside the central 90% of the width and above the bottom 5%. In 9:16 at 1080 by 1920, treat the bottom fifth of the frame as unavailable and the top tenth as risky, which leaves roughly the middle 70% of the height. Then check it on a phone, in the actual app, with a real caption running. That test catches more than any specification.

Contrast, size and timing: the legibility rules

Legibility comes down to three measurable things: contrast between text and what is behind it, the size of the text, and how long each caption stays on screen.

WCAG asks for a contrast ratio of at least 4.5 to 1 for normal text and 3 to 1 for large text, where large means roughly 24 pixels or 18 point. Those numbers are worth applying to the captions and to the text inside the animation itself. The arithmetic is simple: if you know the two colours, the ratio is measurable, and you can check it before the film is rendered.

Size works the same way. Type set at 24 pixels inside a 1080 pixel tall frame looks fine on a laptop and vanishes on a phone held at arm's length, so accessible animation needs a floor rather than a preference. Around 30 to 40 pixels in a 1080 frame is a workable minimum.

Then there is the thing most caption files get wrong: reading speed. Around 160 words per minute is comfortable for a general audience, which works out at roughly two and a half words per second, or about 40 characters per line with two lines on screen at once at most. Anything faster and the viewer is reading instead of watching.

Three more rules that hold in practice. Put a nearly solid plate behind the captions instead of a thin outline, because outlines disappear over pale footage. Keep captions out of the same band as the film's own on-screen text, or the two read as one broken sentence. And hold each caption long enough for a slow reader.

One small advantage worth mentioning: Nine draws the words into the picture, in your own typeface when its licence allows it, so the text in the film looks the same everywhere it plays. Captions should be set in a deliberate caption face rather than whatever the player defaults to, and if you want that styling to match the film, it has to be requested rather than assumed.

Audio description: the part most people forget

If your film carries information in the picture, captions are only half the job. A viewer who cannot see the film needs to be told what is happening, and that is audio description: a narration that describes the visual information in the gaps between the dialogue.

It sounds expensive and it is not, if it is designed in. If the film relies on charts, diagrams or motion to make its point, commission a described version as a second audio track and make sure the script leaves room for it. The worst outcome is discovering at delivery that the script has no spare words for a description the tender promised.

Ask any supplier two questions before you start: can you deliver an audio-described version, and does the price include it? The second question is the one that produces useful answers.

Transcripts are an SEO asset, not a courtesy

A transcript is the least glamorous deliverable and the one that pays back the most. It makes the film available to people who cannot hear it, and it makes the content available to everything that reads text rather than watches video, which includes search engines.

A video with no transcript is a page with no words on it. The words your film says are the words people search for, so put them on the page next to the film, in HTML, with a proper heading structure. A transcript that only exists as a PDF is a transcript nobody reads.

Two ways to get more out of the same work. Write the transcript from the script rather than from an automatic transcription, so names, products and numbers are spelled correctly. And keep it on the page with the film, because a reader who lands on the transcript has arrived at your film and should be able to press play from there.

A checklist to hand a supplier

Give this to whoever makes the film, before the price is agreed. The answers tell you a great deal about how the supplier works.

1. Captions in the film's own language, delivered as SRT and WebVTT files, not burnt into the picture.
2. Subtitle files for each target language, with the speaker identified when a speaker changes.
3. A full transcript as HTML on the page where the film is published, written from the script, not machine-generated.
4. An audio-described version if the film carries information the narration does not speak.
5. Caption safe areas confirmed for 16:9, 9:16, 1:1 and 4:5, decided at the composition stage rather than after approval.
6. Contrast ratios for the film's own on-screen text and for the caption style, stated as numbers.
7. Reading speed for the caption file, stated in words per minute, and a maximum of two lines on screen at once.
8. A statement of what you will receive if the film is revised: refreshed caption files, refreshed transcript, and the date they were refreshed.

None of that requires a specialist accessibility consultant. It requires the film to be planned on the assumption that it will be read as well as heard.

If you want to see how your own colours behave on a moving screen, Nine's free page reads your website and plays Nine's own films in your colours, with no account. A film of your own comes with a plan, and how it works is on the front page.

Two articles worth reading next: where each aspect ratio actually goes, which is the other half of the safe area problem, and animation for training and procedures, where a film nobody can hear is a procedure nobody followed.