Every feed that matters autoplays muted. TikTok, Reels, Shorts, X, LinkedIn — you can turn the sound on, and most people don't, because they're on a train or in an office or next to someone asleep. The figure most often quoted is that around 85% of feed video is watched silently. It comes from platform research rather than from us, and the exact number matters less than the shape of it: silent is the default, not the exception.
Which means the first thing your video does is play to someone who cannot hear it.
The first three seconds do the work
A viewer scrolling gives you a beat to answer one question — is this for me? With sound, your opening line answers it. Without sound, a face moves and no information arrives, and the thumb keeps going.
Captions answer the question in the only channel that's actually open. Not as an accessibility afterthought bolted to the end of the pipeline, but as the thing carrying your premise during the window where the decision gets made.
A silent video without captions isn't a quieter version of your video. For most of the audience it's a different video, and a worse one.
Why the burned-in kind
Platforms will generate captions for you, and there are three reasons we don't think that's the same thing.
- They're off by default in most places. A viewer has to opt in to something they don't know exists.
- You don't control the timing or the look. The platform decides when text appears and how it's styled, which is fine until it lands over your product.
- They don't travel. Repost the file elsewhere and the captions stay behind, because they were never part of the video.
Burning captions into the frame means what you see in the editor is what every viewer sees, on every platform, whether or not they touched a setting.
The part people get wrong
The instinct is to caption the whole video evenly, like a transcript. That's correct for accessibility and wrong for retention, and you want both.
What actually helps is treating the opening differently: fewer words on screen at once, larger, positioned where the platform's own interface won't cover them. Three words a line is the default here for exactly that reason — it's readable at a glance in a way a full sentence isn't, at the moment when a glance is all you get.
After the first few seconds you can relax. The viewer has decided to stay, and the captions go back to doing the ordinary job of letting someone follow along.
What it costs to try
Roughly a minute of your time. Upload, let it transcribe, pick a look. If you have any analytics on your own posts, the honest test is to caption half of what you publish for a fortnight and compare — your audience, your content, your numbers, rather than anyone's blog post including this one.
We'd rather you did that than take our word for it.