Burned-in captions

Captions that are part of the video itself, so they show on every phone and app and nobody can turn them off.

Caption text rendered into the video frames at export (open captions), as opposed to a timed text track such as SRT or WebVTT that the player renders on request (closed captions).

A burned-in caption is pixels. Once the video is exported, the words are part of every frame, like the colour of your shirt, and the file carries no text a player could read or hide.

A caption file is text with timings: this line from 0:03.2 to 0:05.8, the next from 0:05.8 to 0:07.1. The player draws it over the video when the viewer asks, in the player's own font. That makes it searchable and easy to translate, but how it looks is up to the app, not you.

Short-video creators mostly want the first kind: the look is yours, and it is the same on every phone.

A viewer scrolling a feed decides in a second or two whether to stay, and plenty of them have the sound off. Captions in the picture carry the hook to them anyway. Burned-in captions also let you style the words as part of the edit: a keyword in a brand colour, an emoji beside the claim, one word at a time to pace a fast line. A caption file can't do any of that, because the player chooses how it looks.

• You can't fix them after export. A wrong word is in the picture. Fix it in the editor and export again.
• They can clash with an app's own text. A platform's buttons, handle and description sit over the bottom of the frame, so captions placed too low get covered.
• They can double up. If a platform adds its own automatic captions on top, the viewer sees two sets. Keep a copy without captions for those.
• They aren't text. A search engine or a screen reader can't read words in pixels, so the spoken words and the description still matter.

Short lines read faster than long ones: a phrase or a few words at a time, not a full sentence. Keep them clear of your face and of the bottom strip where the app's buttons sit. Use a bold font with an outline, a shadow or a plate behind it, so the words stay readable over a bright window or a busy shirt. Mark only the words that matter, because when every word is coloured, none of them stands out. Cut the ums from the captions even when they stay in the sound, so the reading is cleaner than the speaking.

In a video where you talk to the camera, the captions are your words, so they change every time the cut changes. When a jump cut drops a repeated line, its words must go too, or the captions run ahead of you. That is why captions are best made from the final cut rather than typed in before it, and why an editor that cuts by your words can rewrite them as it cuts. A punch-in moves your face in the frame, so check the captions still sit clear of it.

Turn on Captions in Auto Edit, and Bang AI writes word-by-word captions from what you said in the cut it makes. In the editor you pick a look, such as Pop, Karaoke, Plate or Subtitle, fix a word, hide ums and uhs, and mark keywords and emoji. The captions are burned into the video you export; Bang AI doesn't make SRT or VTT files. On iPhone and the Mac, Export Pack makes each shape twice, with captions and without, so you have a clean copy for a platform that adds its own. Captions walks through every step.