For widescreen video, the working standard is at most 42 characters per line and two lines per subtitle, shown at no more than about 20 characters per second, and on screen for between five-sixths of a second and seven seconds. Vertical phone video needs much shorter lines, around 22 characters. These numbers exist so a viewer can read every subtitle and still watch the picture; the rest of this page is what each one means in practice and where it bends.
These are the limits suubbs builds subtitles to. They match the figures large subtitle style guides use for adult programming in English.
| Rule | Widescreen (16:9) | Vertical (9:16) |
|---|---|---|
| Characters per line | 42 | 22 |
| Lines per subtitle | 2 | 3 |
| Reading speed, maximum | 20 characters per second | 20 characters per second |
| Shortest time on screen | 0.833 s (five-sixths of a second) | 0.833 s |
| Longest stretch of speech in one subtitle | 7 s | 7 s |
| A pause that always starts a new subtitle | 0.6 s or longer | 0.6 s or longer |
| Longest a subtitle lingers after the speech stops | 0.4 s | 0.4 s |
For reference, Netflix's published timed-text style guide for English uses the same 42 characters and two lines, a reading speed of 20 characters per second for adult programmes (lower, 17, for children's), and a minimum and maximum duration of five-sixths of a second and seven seconds. If you deliver to a broadcaster or platform with its own guide, its numbers win over anyone's defaults, including these.
The limit is about the eye, not the screen: a line has to be readable in one or two glances.
Count every character, spaces and punctuation included. "I don't think that's what she meant." is 36. At 42 a line spans a comfortable part of a 16:9 frame at normal caption size, and two of them hold about fourteen or fifteen words, which is as much as most viewers can take in while still looking at the picture.
Vertical video breaks that. A 9:16 frame is about half as wide as it is tall, so at the type sizes people use on phones, roughly twenty characters fill the width. That is why suubbs' vertical layout allows 22 characters a line and lets a subtitle use three lines rather than two: the same amount of text, stacked taller. If a line is wider than the frame allows, the player wraps it somewhere you did not choose, often leaving a single word alone on a line.
Short-form captions often go further and show only one to four words at a time. That is a different style with different rules, covered in word-by-word captions.
Line length is easy to see. Reading speed is not, and it is what makes subtitles feel rushed.
Reading speed is the subtitle's total characters divided by how long it is on screen. A two-line subtitle of 80 characters needs four seconds at 20 characters per second. Give it two and a half and most viewers will not finish it before it disappears.
Ordinary conversational English, at very roughly 150 words a minute, comes out at around 15 characters a second, under the limit, so for most speech the limit is met just by showing each line while it is spoken. It bites on fast talkers, on dense technical speech, and on a short subtitle squeezed between two others. There are three fixes, in order of preference:
The lower limit matters too. A subtitle on screen for less than about five-sixths of a second (20 frames at 24 fps) is a flash: people notice it changed and cannot read it. Very short lines get extended to that minimum, as far as the pause after them allows.
Two subtitles with the same words can read very differently depending on where they break.
Between subtitles, break at the end of a sentence first, then at a clause (after a comma, semicolon, colon or dash), and always at a real pause in the speech. A subtitle that runs across a long pause looks frozen while nobody speaks. suubbs starts a new subtitle at any pause of 0.6 s or more, and never lets one cover more than seven seconds of speech.
Between the two lines of one subtitle, keep words that belong together on the same line. The common rule is never to end a line on a small word that leans on the next one:
She said she would meet us at the
station before the last train left.
She said she would meet us
at the station before the last train left.
The second is easier to read even though its lines are uneven. suubbs applies this when it breaks lines: it will not end a line on an article, a short preposition, a conjunction or a possessive (a, the, of, to, and, but, your, their, and short verbs like is and was), and steps the break back a word or two to avoid it.
Leave a gap. When one subtitle ends and the next begins at the same instant, the change can be missed and the two read as one. A frame or two of space between them makes each change visible; suubbs leaves up to 80 ms.
Nobody checks characters per second by hand on 600 subtitles, and nobody should have to.
suubbs transcribes the video on your own Mac and gives every word its own timestamp, then builds subtitles from those words using the limits above, so the first draft already meets them wherever the speech allows. Where it cannot, because someone talks faster than 20 characters a second with no pause to borrow, the line list marks that subtitle, and hovering the mark says why: "Reads at 24 characters per second, over the 20 limit", or "A line here runs past 42 characters, so it will wrap awkwardly".
Those are the lines to rebalance or condense by hand. Split a subtitle at the playhead
with S, merge it with the next with M, or retype it. To change
the layout of the whole file, pick Widescreen or Vertical /
social under Line layout in Timing & layout and the
subtitles re-split to it straight away (Re-split lines does the same on
demand). Re-splitting rebuilds the subtitles from the words and
leaves any line you have edited or locked exactly as it is, so the hand fixes survive.
Subtitles you import as an SRT or VTT count as yours and are left alone too. One thing
to expect with an import: a subtitle that was on two lines arrives joined onto one, so
it is flagged as too wide until you Alt-click the word that should start the second
line.
The limits themselves are fixed. You can switch between the two layouts and cap how many words appear at once, but not set, say, 17 characters per second for a children's programme; a guide with stricter numbers than these has to be met by editing.
One exception is deliberate. With Words on screen set to four or fewer, for word-at-a-time social captions, the reading-speed and minimum-duration rules are switched off: each group follows the speech exactly, because holding it longer would push it into the next.
The free browser version does all of this on the first 60 seconds of a video, with no account. The Mac app, $9.99 once, does the whole video with a larger model. Export as SRT, VTT or ASS; if a platform wants the other web format, converting SRT to VTT is two small changes.
The short answers, with the numbers.
At most 42 for widescreen video, over no more than two lines. For vertical phone video, keep lines to around 22 characters, since the frame is only about half as wide.
No more than about 20 characters per second for adult viewers, counting spaces and punctuation. Children's programming commonly uses a lower limit, such as 17.
Long enough to read at the reading-speed limit, at least five-sixths of a second, and no longer than about seven seconds. Do not keep it up long after the speaker stops; a fraction of a second is enough.
Yes. Line length and reading speed both count every character on screen, including spaces and punctuation.
It depends on the job. Subtitles for reading are often condensed to meet the reading speed; captions for accessibility or transcripts for the record should match what was said. Decide before editing, not line by line.
Yes. Import the SRT or VTT over its video and the line list marks any subtitle over the reading-speed or line-length limit. Imported subtitles are kept as they are, so the fixes are yours to make. A two-line subtitle arrives joined onto one line, so put its break back first or it will read as too wide.