You’ve got captions on your video, and something feels off. Either the words flash past before anyone can read them, or a long line sits there while the speaker has already moved on. Both come down to the same question: how long each caption stays on screen, and what decides it.
The short answer. Traditional subtitles stay up for roughly one to seven seconds and are limited to about 20 characters per second of reading. Short-form captions work differently: they follow the speech word by word, so each chunk stays up exactly as long as it takes to say it. Keep chunks to one to three words, never let one vanish faster than it can be read, and hold the last line a beat after the voice stops.
Where the broadcast rules come from
Film and TV subtitles have had decades of style guides, and they mostly agree. Netflix’s English Timed Text Style Guide sets a minimum of five-sixths of a second and a maximum of seven seconds per subtitle, with up to 42 characters on each of two lines. It caps reading speed at 20 characters per second for adult programmes. The BBC’s subtitle guidelines put their target at around 160 to 180 words per minute.
Those numbers exist because broadcast subtitles are an edited reading of the dialogue, not a transcript. A subtitler often condenses a fast speaker so the text stays readable, then times each block so the viewer has enough seconds to take in two full lines. The viewer reads, glances at the picture, and reads again. Duration is a reading problem.
Why short-form captions run faster
Captions on TikTok, Reels and Shorts usually break every one of those rules. A single word might sit on screen for a fifth of a second. That still works, for three reasons.
The text is verbatim and in sync
Word-by-word captions show what’s being said as it’s said. Your eye isn’t reading ahead of the audio or catching up to it. Even on mute, the mouth movements and the rhythm of the edit tell you when the next word is coming, so a short duration doesn’t feel rushed.
The chunks are tiny
Reading two lines of 42 characters takes real time. Reading one word takes almost none, because you recognise a short word as a shape rather than spelling it out. The faster the chunks change, the smaller they need to be. That’s why fast, punchy edits tend to use one word at a time, while calmer talking-head videos can carry a short phrase.
The viewer can loop
A short video replays on its own. Nobody expects to catch every word of a 20-second clip on the first pass the way they would in a two-hour film. That’s not an excuse for unreadable captions, but it does change the trade-off.
So the speaker sets the timing. What you control is how much text each chunk holds, and what happens at the edges: the start, the pauses and the end.
How to check your own timing
You don’t need to count frames. A few quick checks catch nearly every problem.
Use characters per second as a ceiling
Netflix’s 20 characters per second is a decent upper limit for any chunk. Count the characters in a chunk, spaces included, and divide by 20 to get the shortest time it should be visible. A three-word chunk like “this changed everything” is 23 characters, so it wants a little over a second. If the speaker says it in half a second, that chunk is too big. Split it into “this changed” and “everything”, or show one word at a time.
Watch it on mute, at full speed
Turn the sound off and watch the whole clip once without pausing. If you have to replay a moment to catch what it said, that part is either too fast or too long for its slot. Then watch it on a phone held at arm’s length, the way people actually see it. Text that reads fine on a laptop can be too small to scan in a fifth of a second on a phone.
Look for flicker
Flicker happens when a caption disappears for a split second between phrases and then the next one appears. It looks like a glitch and makes the eye hunt for where the text went. If there’s a gap shorter than about half a second, let the previous chunk stay up until the next one starts.
Pauses, cuts and the last line
The middle of a clip mostly takes care of itself. The edges need a bit of attention.
- The first caption. Have it appear on the first spoken word, not before. Text that lands before the voice reads as a title, not a caption.
- Long pauses. If the speaker stops for two seconds or more, or you cut to b-roll with no voice, clear the screen. Old words sitting over a new shot read as a mistake.
- Hard cuts. Where a cut lands mid-phrase, try to start a new chunk on the cut rather than carrying half a phrase across it.
- The last line. Hold it for a moment after the voice stops. A final word that disappears on the last frame is the one most viewers miss.
When the timing itself is wrong
Everything above assumes the captions are in sync with the speech. If whole phrases appear early or late, no amount of chunking will fix it. That’s a timing problem in how the captions were made.
Hand-typed captions are the usual cause, because they get timed by eye in blocks. Auto captions that come back with word-level timing avoid most of this, since every word carries its own start and end time. Capster works that way, which is what lets the animated styles highlight or pop each word exactly as it’s spoken. The steps to add captions to a video on iPhone cover the full flow from import to export.
Chunk size is a style decision, so it lives with the rest of the look. The animated caption styles page shows which layouts show one word, a short phrase or two lines. In Capster, One Word suits fast speakers and the two-line styles suit slower, conversational ones.
Once you’re happy with the timing, burn the subtitles into the video so it plays the same way on every platform. If they look soft after you upload, the fix is covered in why captions look blurry after upload.
A quick timing checklist
| Check | Rule of thumb |
|---|---|
| Chunk size | One to three words for most short-form video |
| Fastest a chunk can go | Its character count divided by 20, in seconds |
| Gaps under half a second | Hold the previous chunk instead of blanking |
| Silence over two seconds | Clear the screen |
| First caption | Appears on the first spoken word |
| Last caption | Stays up a beat after the voice stops |