Add captions to a video

The sound is off, and that changes the job

Short video is watched in places where sound is rude. A commute, an office, a bed with somebody asleep next to you, a queue. The clip has to work as a silent thing first and a heard thing second, and captions are what makes that possible. This is not a nicety bolted on for accessibility, although it is that too and that alone would be reason enough. It is the difference between a clip that plays and a clip that gets scrolled past.

It also changes what good captions look like. Subtitles for a film are designed to be unobtrusive, sitting quietly at the bottom, waiting to be read when needed. Captions for a Short are the opposite: they are a main element of the composition, they carry the pacing, and they are frequently the only thing moving on screen. Treating one like the other is why so many captioned Shorts feel flat.

Word by word, and the small lie that makes it feel right

A caption that appears a whole line at a time gives away the ending. The eye reads faster than the mouth speaks, so the viewer finishes the sentence before the speaker does and then waits, which is exactly the moment attention leaves. Lighting one word at a time keeps the reading speed and the speaking speed together, and the clip holds.

There is a small deliberate inaccuracy in how that is done here and it is worth explaining. A word lit at the exact millisecond the transcriber says it starts reads as late, because the viewer has already heard the beginning of it. So the highlight leads very slightly, by an amount small enough to still feel word accurate and large enough to stop feeling behind. Only colour and the box opacity ramp during that lead, never the size, because growing the word would shift every word after it on the line and the whole caption would jitter.

Individual words can also be emphasised, which is what turns a caption from a transcript into a piece of editing. There is an automatic pass that marks the words held longest and landing loudest, a few per clip and never two in a row, and you can add or remove them by hand.

The interface is drawn on top of your video

Every platform draws its own things over the picture. A caption placed at the very bottom of the frame, which is where a film would put it, ends up underneath a username, a description, a row of buttons, or a progress bar, depending on where it is playing. The safe area is narrower than the frame, and the practical answer is to keep captions comfortably above the bottom third and inside the sides.

The same applies at the top, where a header and a back button live. Captions here default to a position that clears both, and the position is adjustable, because the safe area is not identical on every platform and the clip you are making might only ever be going to one of them.

Contrast is the other half of readability, and the reason the caption styles here include a plate and an outline rather than only plain text. Plain white text over a bright, busy or moving background is unreadable at exactly the moments you most want it read. An outline keeps the letterforms legible against anything, and the stroke is painted behind the fill rather than over it, so thin faces do not get eaten from the inside.

Doing it without an editor open

None of this needs a timeline. Paste a link, let it transcribe, fix anything it misheard by retyping the line, pick a style, and export. The whole thing runs in a browser, which means it also runs on a phone, which is where a fair amount of this work actually gets done.

If you also want a separate subtitle file for a long upload elsewhere, that comes out of the same transcript as an SRT or VTT, so the burned in captions and the file cannot say different things.

How to do it

Specifications

The numbers rather than the adjectives, including the limits. If one of these is a dealbreaker it is better learned here than after paying.

What people ask about this

Keep going