On short-form platforms, captions are not optional. A large share of people scroll with the sound off, and the first second decides whether they stay. If they cannot follow what you are saying at a glance, they are gone. Good captions keep them watching; bad captions — tiny, cluttered, covered by buttons — are almost as bad as none.
Should you burn captions in or use a subtitle file?#
Burn them in. A burned-in caption is part of the video image, so it looks identical on TikTok, Instagram Reels, YouTube Shorts and X, and it survives downloads and re-uploads. Subtitle files depend on each player, are often off by default, and cannot be styled to match your brand.
How big should captions be?#
Bigger than you think. People watch on a phone at arm's length, often while moving. A few rules that hold up:
- Large, bold type with a simple sans-serif or a strong display face.
- High contrast: light text with a dark outline or shadow, or the reverse, so it reads over any background.
- No thin weights. They disappear over busy footage.
Where should captions go on a vertical video?#
Every app places its own interface over your video: the account name and description along the bottom, and buttons down the right side. Captions placed at the very bottom get covered. Put them in the middle-to-lower third, below the speaker's face and clear of the edges.
Framing matters here too. If the speaker drifts to the edge of a vertical crop, captions end up over their face. A crop that keeps the speaker centred keeps the face and the captions in their own zones.
How many words should be on screen?#
A short phrase at a time. Long sentences wrapped over two lines turn a video into reading homework. Break captions at natural pauses so each chunk is a unit of meaning, not an arbitrary cut mid-phrase.
| Choice | Works | Avoid |
|---|---|---|
| Words per chunk | A short phrase | Full sentences over two lines |
| Position | Middle-to-lower third | The bottom edge, under the app's text |
| Style | Bold, outlined, high contrast | Thin type with no outline |
| Timing | Matches each word as it is said | Whole lines that appear early or linger |
| Emphasis | Highlight the current word | Random colors and emoji on every line |
Does highlighting the spoken word help?#
Yes. Highlighting the word being spoken — in your brand color, for example — gives the eye an anchor and makes the clip feel faster. Use one highlight style consistently; changing colors on every line becomes noise.
How accurate are AI-generated captions?#
Modern speech recognition handles everyday speech well, but it still trips on names, product terms, jargon and fast cross-talk. Read every caption once before you post. In OpenCMO you can fix a word directly in the transcript, and the captions update with it.
How do you keep captions on brand?#
Pick one caption style — font, colors, highlight — and use it on every clip so people recognise your videos before they read your name. OpenCMO's brand kit stores that style once and applies it to a clip in one step, so you do not restyle captions by hand each time.
The checklist: burn them in, make them big and high-contrast, keep them in the middle-to-lower third, show a few words at a time, sync them to the speech, and proofread once. If you are using clips to grow a product, see short-form video for founders next.