Burned-in captions or a subtitle file?

Captions are not one feature. There are two different things called subtitles, they fail in different ways, and picking the wrong one is a common reason a clip that read perfectly on your screen is unreadable in someone's feed.

The two kinds

Burned-in captions are drawn into the video's pixels at export. They are part of the picture: nothing can turn them off, restyle them, or fail to load them. Every platform shows them identically because there is nothing to show: they are the video.

A subtitle file (.srt, .vtt) is a separate track of text and timings that the player renders. It can be toggled, translated, restyled by the viewer, and read by anything that indexes text.

When to burn them in

Short-form video, almost always. Three reasons, in order of how often they decide it:

1. Sound-off viewing is the default. A caption the viewer has to enable is a caption most viewers never see, and the first two seconds, where the decision to keep watching happens, are exactly where a toggled-off track costs the most.
2. Styling is part of the edit. Word-level timing, a colour on the emphasised word, a font that matches the channel: none of that survives a subtitle file, because the player owns the rendering.
3. Consistency across platforms. Each platform renders subtitle tracks its own way, at its own size, in its own position, sometimes behind the interface chrome. Burned-in captions look the same everywhere because the platform has no say in them.

When a subtitle file is the right answer

• Long-form video on YouTube. The track is indexable, drives search inside the platform, and lets viewers turn it off. A 40-minute talk with permanently burned-in captions is a worse viewing experience, not a better one.
• Accessibility requirements with a specification behind them. Some contexts require a real caption track rather than pixels, and pixels cannot satisfy them.
• Anything that will be translated later. Text stays text; a burned-in caption is a re-export in every language.

The two are not exclusive. The common professional setup is burned-in captions on the short vertical cut and a subtitle file on the long horizontal original, from the same transcript.

The part people get wrong

Safe area. Every feed puts interface over the video: a caption sitting in the bottom 15% of a vertical frame is behind the username, the description and the buttons on at least one platform. It looked fine in the editor because the editor does not draw someone else's interface over your work.

Keep captions in the middle third vertically, check the clip in the platform's own preview before publishing, and treat "it looked right in the export" as an untested claim.

What MakeRoll does

Captions are generated from the transcript with a timestamp on every word, styled in the editor, and burned into the exported video, which is what short vertical clips want. There is no separate .srt export today. If that is a blocker for your workflow, it is worth telling us: it is a small feature with a clear shape, and the demand for it is what decides when it gets built.