Write subtitles that remain readable on a phone

Automatic captions can produce plausible words that are wrong in context. Names, technical terms and short negatives deserve particular attention because a small error can reverse the meaning. Review captions against the recording rather than treating a smooth-looking transcript as proof of accuracy.

Edit in units of meaning

Break lines where the sentence naturally divides. Avoid separating a person's first and last name, a number from its unit or a short phrase that needs to be understood together. The viewer should be able to read a caption as a meaningful piece of speech rather than reconstructing it from awkward fragments.

Do not remove a qualification simply to fit more words onto the screen. If the speaker says a feature applies only to a particular version, that boundary matters. Shorten surrounding repetition where appropriate, or adjust the timing and edit, instead of converting a limited claim into a general one. When exact wording is important, preserve it carefully.

Check the image and text together

Use clear contrast and a background treatment that remains readable across changing footage. Place captions where they do not cover the action being explained, a speaker's mouth when it matters, or an essential screen control. Preview the exported video in its intended display size; a design that looks comfortable on a large monitor may be difficult on a phone.

Watch once without sound. Can the viewer understand the important spoken information and identify changes of speaker where needed? Then watch with sound to check synchronisation and omissions. Include meaningful non-speech information when it is necessary to understand what is happening, rather than adding decorative descriptions of every background noise.

Keep versions manageable

Maintain a corrected caption file or transcript alongside the final video so later edits do not begin from the uncorrected automatic version. If the spoken edit changes, review the captions again. A subtitle track from an earlier cut can be accurate word by word yet appear at the wrong moment.

W3C's caption guidance distinguishes captions that convey speech and relevant non-speech audio from a transcript that is not synchronised with the video. That distinction matters in production: a text file below a player can be useful, but it does not perform the same job as words appearing at the relevant moment. Check the intended viewing experience rather than treating every text alternative as interchangeable.

Try a short caption-quality pass on a difficult segment rather than only watching the easy opening. Choose the fastest speech, a specialist name and a moment with important visual action. If those sections fail, adjust the edit or caption layout before checking the rest. This targets the places where a polished automatic output is most likely to conceal a practical reading problem.

Avoid claiming that adding captions alone makes the entire video accessible. Other information may exist only in the visuals, and the viewing context can introduce further barriers. Treat readable captions as one part of making the explanation usable. The immediate goal is clear: people should not have to pause repeatedly or guess a misrecognised word to follow the central point.