Captions are readable when each one stays on screen long enough for a typical viewer to read it comfortably while still watching the picture. In practice that means keeping text to a sensible number of characters per second (CPS), limiting each caption to one or two short lines, and giving every caption a minimum time on screen. The exact figures vary between style guides, but the principles are the same everywhere.
Why reading speed matters
A caption that disappears before it has been read is worse than no caption: the viewer loses the words and misses the picture while trying to catch up. Fast speech makes this easy to get wrong, because automatic captioning tools time text to when words are spoken, not to how long they take to read. A presenter who talks at 220 words per minute produces captions that are technically accurate and practically unreadable.
Reading speed also affects people you might not be thinking about: d/Deaf viewers who rely on captions entirely, viewers reading in a second language, children, and anyone watching on a small phone screen with the sound off. Our guide to captions and accessibility for UK creators covers the wider picture.
The three ways to measure speed
Style guides express reading speed in slightly different units. It helps to know how they relate.
| Measure | What it counts | Typical use |
|---|---|---|
| Words per minute (WPM) | Words shown per minute of caption time | Broadcast guidelines, including the BBC's |
| Characters per second (CPS) | Characters (usually including spaces) per second on screen | Subtitle editors, streaming style guides |
| Seconds per word | Minimum display time per word | Quick manual timing checks |
As a rough conversion for English, an average word plus the following space is about six characters. So:
- 150 WPM ≈ 2.5 words per second ≈ 15 CPS
- 180 WPM ≈ 3 words per second ≈ 18 CPS
Treat those as approximations. Short, punchy words push the character count down; technical vocabulary pushes it up. This is why many editors prefer CPS: it measures what the eye actually has to read.
One trap: tools disagree on what counts as a character. Some include spaces and punctuation, some exclude them, and some ignore line breaks. A caption that shows 17 CPS in one program might show 15 in another. Pick one tool for checking and stick to it for the whole project.
What the BBC guidelines say
The BBC publishes its subtitle guidelines openly at >bbc.co.uk/accessibility/forproducts/guides/subtitles, and they are one of the most useful free references for anyone captioning in British English. Figures in them are revised from time to time, so read the current version rather than relying on a summary, but at the time of writing the broad approach is:
- Speed: for pre-recorded content, subtitles are timed to a rate of roughly 160 to 180 words per minute, with slower rates appropriate in some situations, such as content aimed at young children.
- Follow the speech: subtitles should match the spoken words as closely as possible, and editing text down is something to do only when needed to keep the speed readable, not as a default.
- Line length and count: subtitles are normally one or two lines, with line length limited so text does not stretch across the full width of the screen. The traditional broadcast figure is around 37 characters per line, and the guidelines also describe limits in terms of screen width for online players.
- Line breaks: break lines at natural linguistic points, so that phrases are not split awkwardly across lines or across consecutive subtitles.
- Synchronisation: subtitles should appear and disappear in step with the speech and respect shot changes where possible.
Streaming services publish their own style guides with figures in the same general range. Several use a maximum of about 42 characters per line and an adult reading speed of around 17 to 20 CPS for English, with lower limits for children's content. If you are delivering to a specific broadcaster or platform, their specification always overrides general advice.
A practical house standard
If you have no client specification, the following is a sensible, conservative starting point for YouTube and general online video. It sits comfortably inside the ranges above.
| Setting | Recommended starting point |
|---|---|
| Maximum reading speed | 17 CPS (up to about 20 for adult, fast-paced content) |
| Maximum line length | 42 characters (37 if you want a broadcast feel) |
| Maximum lines per caption | 2 |
| Minimum duration | About 1 second, even for a single word |
| Maximum duration | About 6 to 7 seconds; split longer captions |
| Gap between consecutive captions | A couple of frames, so the change is noticeable |
Short-form vertical video is a special case: captions there are usually one short line of a few words, and the rules shift accordingly. See short-form captions for that format.
Worked example: fixing a caption that is too fast
Here is a caption exported from automatic transcription:
14
00:02:10,400 --> 00:02:12,300
So the thing you really need to remember is that the frame rate
has to match the project settings exactly.
Count the characters: the two lines total about 105 characters including spaces. The caption lasts 1.9 seconds. That is roughly 55 CPS, around three times too fast. The first line is also well over 42 characters.
There are three ways to fix it, and good captioning uses them in this order.
1. Extend the timing, if there is room. If the next caption does not start until 00:02:17, the end time can be pushed out. At 17 CPS, 105 characters needs about 6.2 seconds.
2. Split it into two captions, each timed to its part of the speech:
14
00:02:10,400 --> 00:02:14,000
So the thing you really need
to remember is that the frame rate
15
00:02:14,100 --> 00:02:16,900
has to match the project
settings exactly.
Caption 14 is now 63 characters over 3.6 seconds (about 17.5 CPS); caption 15 is 42 characters over 2.8 seconds (15 CPS). The speech has to allow this, of course: a split only works if the words are actually spoken across that time.
3. Edit the text down, carefully. If the speaker genuinely says all of it in under two seconds and the next caption follows immediately, you may need to condense: "Remember: the frame rate must match your project settings exactly." Keep the meaning and the speaker's voice. Do not remove words simply to hit a number when timing changes would solve the problem.
Checking a whole file quickly
Checking every caption by hand is slow. A short Python script can flag the ones over your limit:
import re
LIMIT_CPS = 17
MIN_DUR = 1.0
def secs(t):
h, m, s = t.replace(",", ".").split(":")
return int(h) * 3600 + int(m) * 60 + float(s)
text = open("captions.srt", encoding="utf-8-sig").read()
for block in re.split(r"\n\s*\n", text.strip()):
lines = block.splitlines()
if len(lines) < 3 or "-->" not in lines[1]:
continue
start, end = [secs(x.strip()) for x in lines[1].split("-->")]
body = " ".join(lines[2:])
body = re.sub(r"<[^>]+>|\{[^}]*\}", "", body) # strip tags
dur = end - start
cps = len(body) / dur if dur > 0 else 999
if cps > LIMIT_CPS or dur < MIN_DUR:
print(f"#{lines[0]} {dur:.2f}s {cps:.1f} CPS {body[:50]}")
It counts spaces and punctuation, which is the more conservative convention. If you would rather not run code, our caption speed checker does the same job in the browser.
Timing habits that improve readability
- Lead in slightly. Many captioners let a caption appear a fraction before the speech starts, because it takes a moment for the eye to move to the text. Do not overdo it; noticeably early captions give away punchlines and feel out of sync.
- Do not flash the next caption instantly. If one caption ends and the next begins on the same frame with similar text, viewers may not notice it changed. A small gap helps.
- Respect shot changes. A caption that hangs a few frames over a cut is distracting. Where possible, end it on the cut or hold it clearly across.
- Break at phrase boundaries. "the frame rate / has to match" reads better than "the frame / rate has to match".
- Do not split a sentence into a fragment that misleads. "I don't think / we should cancel it" is fine; a line that ends on "I think we should cancel" briefly says the opposite.
Timing only helps if the captions are in sync to begin with. If they start correct and slide later over the course of the video, see fixing subtitle sync drift before adjusting reading speed.
Key takeaways
- Aim for roughly 15 to 17 CPS for general audiences, and treat about 20 CPS as an upper limit for adult content.
- The BBC guidelines express speed in words per minute; at the time of writing they suggest roughly 160 to 180 WPM for pre-recorded content. Always check the current version.
- Keep captions to two lines of no more than about 37 to 42 characters.
- Fix fast captions by extending timing first, splitting second and editing text last.
- Use one tool to measure CPS consistently, because tools count characters differently.
FAQ
What is a good characters-per-second rate for YouTube captions?
There is no official YouTube figure. A limit of around 17 CPS is a sensible default, with up to about 20 CPS acceptable for adult, fast-paced content. Slower is better for children's and educational videos.
Should captions be verbatim or edited?
Broadcast guidelines, including the BBC's, generally favour staying close to what was said and editing only when necessary to keep captions readable. Heavy paraphrasing frustrates viewers who can partly lip-read or hear the audio.
Do spaces count as characters?
It depends on the tool. Counting spaces and punctuation is more conservative and more common in subtitle editors. Whatever you choose, apply it consistently across a project.
How long should a one-word caption stay on screen?
Aim for about a second as a minimum. A single word flashed for a quarter of a second is easy to miss entirely, even though it is short.