Most of the advice on captioning a video skips the one decision that determines whether anyone reads them at all: burned-in, or uploaded as a separate file. Get this wrong and the rest of the styling doesn't matter.
Burned-in vs. uploaded: pick burned-in for short-form
An uploaded caption file (.srt or .vtt) is a track the platform can show or hide. On YouTube, viewers have to turn it on. On TikTok, Reels and Shorts, most players don’t support the file at all — the platform either ignores it or auto-generates its own, usually worse, captions from your audio.
Burned-in captions are rendered directly into the video frame. They’re always there, they look exactly the way you designed them, and they work identically on every platform because they’re not a separate feature the player has to support — they’re just pixels.
The tradeoff: burned-in captions can’t be turned off, and if there’s a typo you have to re-render, not just fix a text file. For short-form specifically, that tradeoff is worth it — most short-form video is watched with the sound off, which makes captions the primary way most viewers understand what’s being said, not a backup for the hard of hearing.
Getting the transcript right
Every automatic captioning tool starts from a speech-to-text transcript, and every mistake downstream starts as a mistake in that transcript. Two things to check before you trust it:
- Names and jargon. Speech-to-text guesses at anything it hasn’t seen before — product names, people’s names, technical terms. A wrong guess here reads as a mistake to viewers even though it’s the software’s fault, not yours.
- Punctuation at clause boundaries. Automatic punctuation tends to be roughly right but not exactly right, and a misplaced break changes where the caption line splits — which changes how fast a line reads.
Thirty seconds of proofreading before export catches almost all of this.
Styling for readability, not decoration
- Size: large enough to read on a phone at arm’s length without pausing. If you have to ask “is this too big,” it probably isn’t.
- Contrast: a solid or semi-transparent background block behind the text beats an outline alone — outlines fail against busy or light backgrounds.
- Position: keep captions clear of the bottom fifth of the frame, where TikTok, Reels and Shorts all draw their own UI (like, comment, follow buttons). A caption that’s readable in your editor and covered by an interface element on the actual app is a caption nobody read.
- Line length: one or two short lines per caption, timed to roughly match natural speech pauses. A caption that changes every 0.3 seconds is unreadable; one that sits on screen for four seconds of continuous speech has fallen behind.
Translation and multiple languages
If you’re captioning for an audience that doesn’t share your source language, translate the transcript before styling, not after — translated text runs longer or shorter than the original in almost every language pair, and re-timing captions after they’re already burned in means starting over. Machine translation is good enough for captions in most language pairs now; it’s worth a native-speaker skim before publishing if the content is commercial rather than casual.
Soft Clipper transcribes, punctuates and burns in captions automatically as part of the clipping process, styled to stay clear of platform UI on every format — no separate captioning step or subtitle file to manage.