Creator Library

How to caption short-form video accurately

Build readable captions from a transcript, repair speech-recognition mistakes and check timing, speaker changes and important sounds.

Make a reference list before transcribing

Ask the speaker for spellings of names, product terms and places. Listen to the original and write down any number or unit whose meaning would change if a digit disappeared. In a kitchen demonstration, “one half teaspoon” and “one teaspoon” produce different results. A polished animation cannot repair a wrong quantity.

Break lines where the thought breaks

Keep a name with its title and a number with its unit when space allows. Time each caption to the words that are actually spoken. W3C’s guidance includes meaningful non-speech sounds and speaker identification when needed. If the timer rings off screen and the presenter reacts, a short sound cue can explain the reaction.

  • Correct the transcript against the recording.
  • Play one pass muted and follow the action using the captions.
  • Play a second pass with sound to catch text appearing late or lingering into the next speaker.

Test with the real picture behind the text

A caption that reads clearly against a dark jacket can disappear when the speaker turns toward a window. Scrub across the shot and inspect the brightest frame. Use a backing or move the caption when needed. If the client requests a separate subtitle file, open it with the delivered video to verify synchronization.

Further reading

Related guides