Guide

How long should captions stay on screen?

Long enough to read comfortably — roughly 1 to 6 seconds per line, matched to the audio. Here's the reading-speed rule, and why animated short-form captions play by slightly different math.

The short answer

A caption should stay on screen long enough to read without effort: about 1 second minimum, up to roughly 6 seconds maximum, timed to how long the words are spoken. The classic guideline is a reading speed around 15–17 characters per second. In animated short-form captions, words appear and clear individually, so no single line ever lingers.

The reading-speed rule

Caption timing comes down to one thing: can a viewer finish reading the line before it disappears? The long-standing broadcast convention is about 15–17 characters per second — comfortable for most adults. A short line of 5–7 words fits in roughly 2–3 seconds at that pace.

Two guardrails bracket it: a caption should stay up for at least about a second even if the phrase is tiny (so it doesn't flash), and no longer than about six seconds (past that it feels stuck and the sync drifts from the audio).

Short-form plays by different math

Traditional subtitle timing assumes full sentences on screen. But the word-by-word style that dominates TikTok and Reels works differently: each word appears exactly when it's spoken and clears a beat later. There's no sentence to read — just a word to recognize — so the captions can move as fast as the speech without losing the viewer.

That's why fast-cut captions look almost frantic but still feel easy to follow. The timing isn't about giving you time to read a block; it's about keeping the on-screen word locked to the voice.

Get word-level caption timing synced to your audio automatically.

Try it free

Timing mistakes that cost you viewers

Captions that lag the audio. If the text trails what's being said, it reads as broken. Word-level sync keeps them locked together.
Lines that flash too fast. A full phrase held for under a second can't be read. Give phrase-style lines at least a second.
Lines that overstay. A caption sitting long after the words were spoken feels stuck and makes the video look unedited.
Manual keyframing everything. Timing captions by hand is slow and error-prone; automatic sync gets you 95% of the way in seconds.

How to get the timing right automatically

You shouldn't be counting characters per second on every line. When you upload a clip to AnimateCaptions, it transcribes with word-level timestamps and syncs each caption to exactly when it's spoken — so the timing is right by default. Adjust anything you want, then export a burned-in MP4 with the pacing locked in.

Common questions

Caption timing — FAQ.

How long should a caption stay on screen?

As a rule of thumb, a caption line should stay up long enough to read comfortably — roughly 1 second minimum and about 6 seconds maximum, matched to how long the words are actually spoken. In animated short-form captions, words appear and clear individually as they're said, so no single line lingers.

What is a good subtitle reading speed?

The broadcast convention is around 15–17 characters per second, which most adults can read comfortably. Faster than that and viewers can't finish the line; much slower and it feels sluggish. Good tools time captions to the audio automatically so you rarely have to think in characters-per-second.

Why do TikTok captions change so fast?

Short-form captions are usually synced word by word to the speech, so each word shows exactly when it's spoken and clears right after. It looks fast, but each word only needs to be recognized, not read as a sentence — which keeps pace with energetic delivery.

How do I get caption timing right without adjusting every line?

Use a tool that times captions to the audio automatically. AnimateCaptions transcribes with word-level timing and syncs each caption to when it's spoken, so the timing is right by default — and you can nudge any line before exporting.

Captions timed to your audio, free.

Word-level sync, automatic timing, and a burned-in export. No manual keyframing, no credit card.

Try it free