A Checklist to Run Through Before You Sign Off Your Captions

A caption file can look finished the moment the last line is timed. The text is all there, the video plays, and nothing seems off. That feeling is exactly why a separate review pass matters. While you transcribe and time, your attention sits on getting words down and syncing them. Reading the whole file again, as a viewer would, is a different task and it catches different problems.

I have prepared captions for recorded talks and interviews for over ten years, and the pattern holds on every project. Errors that survive to delivery are rarely dramatic. They are a homophone in the middle of a sentence, a line that flashes by too quickly, a speaker label that changes spelling halfway through. None of these are hard to fix once you see them. The checklist below exists so that you do look.

Start with accuracy, because everything else depends on it. Play the video at normal speed with the captions on and read along without touching the keyboard. Your ear will flag a mismatch faster than your eye will, so resist the urge to read ahead of the audio. Stop only when a line sounds different from what is on screen, and mark the spot.

Pay particular attention to names, numbers and specialist terms. These are the words a viewer is most likely to look up or repeat later, and they are the ones a quick transcription pass gets wrong most often. If the speaker mentions a product code, a place name or a technical phrase, check its spelling against any material you were given. If you were given none, choose one spelling and note it so it stays consistent.

• Names of people, places and organisations spelled the same everywhere
• Numbers, dates and units match what was said
• Specialist terms checked against the supplied glossary or notes
• No words dropped at the start or end of a spoken sentence

Once the words are right, look at how they arrive. A caption can be perfectly accurate and still be hard to read if it appears for too short a time or breaks in an awkward place. During review, watch for lines that vanish before you have finished reading them. If you stumble, a viewer will too, and they cannot pause a live event the way you can pause your editor.

Line breaks deserve the same attention. Try to split a caption at a natural pause, at the end of a clause, rather than between an article and its noun or in the middle of a name. Two short lines are usually easier to take in than one long line that runs the width of the frame. When a caption feels crowded, shorten the wording slightly or divide it into two captions with a clean gap between them.

• Each caption stays on screen long enough to read it once comfortably
• Lines break at natural phrase boundaries
• No more than two lines per caption
• Captions do not linger long after the speech has ended

On recordings with more than one voice, confirm that every speaker change is marked and that each label is spelled and formatted the same way throughout. A label that reads one way in the first minute and another way in the last looks careless and can confuse a viewer about who is talking. Check the moments of overlap especially, where two people speak at once, and decide on a single, steady way to show them.

Sound descriptions need a quick second look as well. Ask of each one whether a viewer who cannot hear the audio would miss something without it. A door closing offstage may not matter, while applause that interrupts a speaker or a phone ringing in the middle of an answer probably does. Keep descriptions short, in the same style each time, and remove any that add nothing to understanding the scene.

Consistency is the quiet quality that makes a caption file feel professional. It covers small decisions that no single line reveals: whether numbers appear as digits or words, how you treat a spoken pause, whether sound descriptions sit in brackets, and how you show a speaker who is off screen. Any one choice is fine. Changing it midway through is what draws attention.

The simplest way to catch drift is to scan the file in two places, once near the beginning and once near the end, and compare how the same kinds of things are written. If you were handed a preferred style by whoever supplied the recording, check against it now rather than after delivery. Fixing a convention in a text editor with a find-and-replace takes a minute, provided you notice before you export.

• Numbers and abbreviations written one way throughout
• Brackets and punctuation for sound notes used identically
• Speaker labels formatted the same from first to last
• Any supplied styling preferences applied everywhere

The last step happens in the finished file rather than the draft. Export the captions in the format you intend to deliver, load them against the video in a player, and watch at least the opening, a stretch from the middle and the closing minute. Problems that hide in an editor, such as lines running past the edge of the frame or timing that has shifted slightly, tend to show themselves here.

If the captions will be embedded in the video rather than supplied as a separate file, check that they are readable against the picture itself. Light text over a bright background can disappear, and a caption sitting where the speaker's name appears on screen will clash with it. A few minutes of looking now spares you a round of corrections once the file is already in someone else's hands.