Video & subtitles · Subtitle Toolkit
Line length, cue duration and reading speed: the rules subtitlers follow
· Why it matters
subtitles accessibility text-processing
Good subtitles are readable, not just accurate. This post explains the conventions professional subtitlers follow for line length, cue duration and gaps between cues, why they exist and how to check a file against them.
The words are right but nobody can read them in time — how formatting decides whether subtitles work
A subtitle file can be accurate, correctly timed and still unreadable. If a line is too long for the time it is on screen, the viewer either misses the end of it or misses the picture while finishing it. Accuracy is a property of the text; readability is a property of the text, its duration and its shape together.
This is the part that automated checking helps least with, which is worth saying plainly at the outset. Structural validation catches cues that cannot display; it does not judge whether a cue that displays correctly can be read.
Line length — why long lines are cut into two and why more than two lines is discouraged
Long lines are split because the eye travels along a line before returning, and a line that spans the frame costs more of that travel than a shorter one. Two lines are the usual maximum, not because three is impossible to read but because a third line occupies enough of the frame to compete with the image and pushes the eye further from the action.
Where the split falls matters as much as whether it happens. Breaking at a grammatical boundary lets the first line be understood on its own; breaking mid-phrase forces the reader to hold an incomplete fragment until the second line resolves it.
Cue duration — minimum time on screen for short lines, maximum before a cue should be split
Duration has a floor and a ceiling for different reasons. Too brief and a cue flashes: it appears and is gone before the eye has settled, which is perceived as a flicker rather than as text. Too long and a short line sits on screen after it has been read and after the speech has moved on, which invites the viewer to reread it.
The degenerate case is caught mechanically. A cue whose start and end are identical has no duration and will not display at all, and validation reports it as a zero-duration cue naming the timestamp. That commonly arises from a clamped shift rather than from authoring, when both ends of a cue were pushed below zero and floored to the same value.
Gaps and overlaps — why cues should not overlap and why a tiny gap between cues helps the eye
Overlaps are a separate fault from ordering. A cue that starts before the previous one has finished is reported as an overlap, naming both the cue and the time the previous one remains on screen, because two cues displayed together is a rendering decision no format specifies consistently. A cue that starts before the previous one began is a stronger fault and is reported as out of order rather than as an overlap.
A small gap between consecutive cues is a convention rather than a requirement. The change from one cue to the next is easier to perceive when there is a brief absence between them than when one is replaced instantly by another, which can read as the first line having simply grown longer.
Reading speed — the idea of characters per second and why it varies by audience and language
Reading speed is the idea that ties length and duration together: characters on screen divided by seconds available. It is the right mental model and the wrong thing to treat as a single number, because the comfortable rate depends on the audience, the language, whether the viewer is also reading other on-screen text, and whether the material is dialogue or dense exposition.
Published guidelines differ for these reasons, and this article does not assert a figure. The durable practice is to compare cues against each other within one programme and against the rate the intended audience is known to manage, rather than to adopt a number from a guideline written for a different context.
Worked example: reviewing ten cues from a short film — identifying which break the conventions and how to fix them
Reviewing a short film, the cues worth attention are the outliers rather than the average. Sort by duration and look at the briefest; sort by text length and look at the longest; the cues that appear in both lists are where a viewer will fall behind. A long line on a brief cue is the classic failure, and it usually arrives from a transcript segmented by sentence rather than by time.
Mechanical faults come out of validation first, which is the cheaper pass: zero-duration cues that will not display, overlapping cues, and cues out of order, each named by number. Fixing those before reading for comfort avoids spending judgement on cues that were never going to appear.
What this does not cover — broadcaster-specific style guides and legal caption requirements
This does not cover broadcaster style guides, which specify line counts, positioning, speaker identification and sound-effect conventions in detail and differ between organisations. Nor does it cover legal caption requirements, which in some jurisdictions mandate the presence and quality of captions for particular content.
It also does not cover the tool enforcing any of this. Validation checks ordering, duration and overlap; it does not measure line length or reading rate, and a file can pass validation completely while being uncomfortable to read.
Takeaway: readable first — how the Subtitle Toolkit's retime and clean features help after the text is right
Get the text right, then make it readable, then check it mechanically. The order matters because retiming a cue whose text is about to change wastes the work, and because a structural fault will mask a readability judgement.
Validation is the fast pass and catches the cues that cannot display at all. Reading comfort is the slow pass and is still a human judgement, made against the audience the subtitles are for rather than against a number.