How to add captions to short videos that are accurate and in sync
Short answer: generate a transcript, correct names and terms, time each word to the audio, keep lines short, place them inside the safe area and check the result frame by frame on a few clips.
Why captions matter
Many people watch short videos without sound, and captions also help viewers who struggle to follow fast speech. Captions are part of the clip, not decoration. Wrong or late captions make the whole clip look careless.
Accuracy first
Automatic transcripts are good but not perfect. Check names, brands, numbers and technical terms. A single wrong word in the first line costs more than a missing word in the middle.
Timing: each word when it is spoken
Word-level timing is what makes captions feel professional. Timestamps from a transcription model can drift into silences, so we align each word to the actual audio. On a real test clip of 142 words we measured the highlighted word changing a median of 16 milliseconds from the spoken word, and 32 milliseconds at worst, by reading the finished video frame by frame.
Line length and style
- Two short lines at most, a few words at a time.
- High contrast against the picture, with an outline or shadow where the background varies.
- One consistent style across all clips from the same show.
Keep captions in the safe area
Platform buttons, names and descriptions cover the top and bottom of vertical video, and screens differ. On a 1080 by 1920 picture we keep text at least 320 pixels from the top and bottom edges and 110 pixels from the sides. Never place captions over a face.
A quick check you can do yourself
- Play the clip once with sound and watch whether each word appears as it is spoken.
- Play it once without sound and check that you understand it fully.
- Step through a few seconds frame by frame at the start and the end.
- Look at the clip on a phone, not only on a monitor.
See it on your own episode
Send one episode and get a free sample from your own recording. No card, no commitment.
Get a free sampleQuestions and answers
How do I make captions match the audio exactly?
Align each word to the audio rather than trusting transcript timestamps alone, and verify on the finished video by stepping through frames.
How many words per caption line?
A few words at a time, in two short lines at most, so it can be read at a glance.
Where should captions go on a vertical video?
Inside the safe area, away from the top and bottom edges where platform interface elements sit, and never over a face.