Subtitles that do not look cheap

Captions are the fastest way to make a video look worse, and almost nobody notices it happening. The preset looks fine on the sentence you were previewing. Then a longer sentence arrives, the text wraps onto four lines, it covers your face, the last line sits under the platform's own buttons, and your video has become somebody's first impression of your business.

Build a preset and test it with short, typical and long caption cues. Recheck it when the format, typeface or platform interface changes, and review every export for exceptions.

One thing outranks all of the styling, so it goes first: a caption's real job is to be accurate and readable. The right words, spelled correctly, present wherever the audio carries meaning, marking who is speaking when more than one person does, and set with enough contrast to be read against whatever sits behind them. That is accessibility, and it is not optional; somebody deaf or hard of hearing, or simply watching muted on a train, is relying on it. Get that right, then make it look like your brand. Do it the other way round and you have a handsome caption that fails the person who needed it most.

What makes captions look cheap?

Not the font, though the font does not help. It is that nothing was tested.

The default caption style in most tools is designed to demonstrate the feature. It sizes itself for a short phrase, it sits wherever the tool decided, and it uses a typeface chosen by somebody who has never seen your videos. It looks acceptable most of the time, and the rest of the time is where the amateurism lives.

Good caption styling survives awkward cases: a long phrase, a busy background and a demonstration near the edge of the frame.

How big should a caption be?

Big enough to read on a phone held at arm's length, and no bigger.

A phone screen shows your caption at a fraction of the size you are designing it at. Preview on a phone before you decide anything, or shrink the preview to the size a phone will show it at. Judging caption size on a laptop at full width is how undersized captions get published. Size is only half of legibility, though; the other half is contrast, so give the text a weight, an outline or a subtle box wherever the picture behind it is busy or pale, and check it against your brightest shot rather than your darkest.

The opposite error is common on vertical video, where enormous captions have become a style. Text that fills a third of the frame leaves nowhere for the picture, and if the picture does not matter you have made a podcast.

Testing at three lengths

This is the whole discipline, and it is one sitting.

The shortest line

Two or three words. Check that it does not look lost, that it sits where every other caption sits rather than floating, and that a background box, if you use one, does not shrink into something odd.

The typical line

A normal spoken sentence, the length most of your captions will be. The preset was probably designed for this one, so it will look fine. Confirm it anyway.

The longest line

Test a long passage, but split it into readable cues rather than fitting a whole spoken sentence on screen at once. Check line breaks, reading time and what the text covers. For many layouts, one or two lines is a useful starting point.

Fix the preset until all three lengths sit in the same place, cover nothing important and never change size between shots. That stability reads as professional more than any styling choice does.

Where do the margins go?

Away from the platform's interface, which is not the same as away from the edge of the frame.

Every vertical platform puts its own furniture over your video: a caption, a username, a row of buttons up the right hand side, a progress bar at the bottom. Your text can be perfectly placed inside the frame and still sit underneath a share button. Open your own last post, look at where the interface falls, and keep your captions clear of it.

Check the current platform interface and the subject in the frame. Keep captions clear of both; a single fixed band will not work for every shot or platform.

Should you use the tool's default typeface?

Use a typeface that stays readable at the delivery size.

Your brand typeface is a good candidate if it stays clear at caption size. Otherwise choose a readable companion face and use it consistently.

If you have none, pick one plain one and use it everywhere. Consistency does most of the work here: a dull typeface used every time beats a fashionable one that changes when a preset does.

The way this studio handles it on short-form work is one preset, styled to the partner's own type law, that holds its margins at every sentence length, tested short, typical and long, and legible before it is brand-styled. It is not a clever idea. It is just a test that was run.

What should a caption never cover?

The thing you are demonstrating.

If your hands are on the keys, the caption does not go over the keys. If you are pointing at the pan, the caption does not go over the pan. This sounds obvious and it is broken constantly, because captions are positioned once at the start and the demonstration moves later. Watch the whole video with the captions on and move any that have wandered onto the subject.

Why does timing make a caption look wrong?

Because a caption that appears before the word is read as an error, not as a caption.

Synchronise caption cues with the corresponding speech and leave enough reading time. Check the full video: timing, wording and cue breaks need attention as well as the preset.

Watch a published video on your phone, both with sound and muted. Check readability, accuracy, timing and whether captions cover the demonstration. Fix the preset where the problem repeats and individual cues where it does not.

Sources and evidence

What here is standards guidance and what is our own method.

Standards and platform facts. The W3C's own description of a caption is a text version of the speech and the meaningful non-speech audio needed to understand a video, accurate and readable enough that somebody relying on it, because they are deaf, hard of hearing or simply watching without sound, gets the same information as somebody who can hear. That is the accessibility floor this piece starts from before any styling decision, and it is why accuracy and legibility are treated as the one part that is not optional, with brand styling built on top of that rather than instead of it.

Our method. Testing a caption preset at three lengths, the shortest line, a typical one and the longest sentence you have ever said on camera, before locking it as a preset you do not revisit, is a workflow test this studio runs, not a standard requiring it.

References.

• W3C Web Accessibility Initiative. Captions/Subtitles. w3.org. Captions include speech and meaningful non-speech audio; closed captions preserve user control. Open captions can still be a sufficient WCAG technique in some contexts.