Split Subtitles by Duration
Or paste subtitle text
Options
A duration — 3000, 3s, 1m30s and 00:00:03,000 all work. Any cue on screen longer than this is divided into as many pieces as it takes, so an 8-second cue with 3000 here becomes three, and the time is shared out evenly whatever the pieces' text lengths are. Use this when the problem is how long a cue stays on screen; use the character box on the other page when the problem is how much text it holds.
Which frame rate the timings are counted in. On the format pages this is what MicroDVD needs, because it stores timing as frame numbers. On the frame rate pages it is the rate the file was timed against — the version the subtitles were made for — and the page's second box takes the rate of the video you are going to use them with. 23.976, 24, 25, 29.97, 30, 50, 59.94 and 60 are the common ones. Getting it wrong shifts the whole timeline, and on the frame rate pages it shifts it by a ratio rather than by a fixed amount.
This page divides any cue that stays on screen longer than your limit into as many pieces as it takes. The limit starts at 7000 ms — seven seconds, the ceiling most style guides quote for a single cue — and if you know your file in seconds rather than characters, this is the page you want. An eight-second cue with the limit at three seconds becomes three cues of about 2667 ms each.
The number of pieces is worked out from the duration, not counted off by hand, and the time is divided evenly between them. The text is then cut at the word boundaries nearest to those even divisions, so the pieces rarely hold exactly the same number of characters. That is deliberate: you asked for durations, so durations are what the tool guarantees, and the text follows as closely as whole words allow.
Nothing moves outside the cue being divided. A cue of exactly the limit is left alone, and a file in which nothing is too long comes back byte for byte — this tool only ever divides, so a limit set too high is a safe first run on a file you do not know. Lower the limit and you get shorter pieces and more of them; the pieces still add up to the original span exactly.
As with every page here the file is read, divided and written again inside this browser tab, with nothing uploaded. SRT, WebVTT, ASS, SSA, SBV, SAMI, MicroDVD, TTML, JSON and CSV are all accepted, and a file whose cue count changed is written in timeline order with numbering from 1.
Examples
An eight-second cue against a three-second limit
Input
1
00:00:00,000 --> 00:00:08,000
The rain kept falling all through the night and no one in the house could sleep.
Output
1
00:00:00,000 --> 00:00:02,667
The rain kept falling all
2
00:00:02,667 --> 00:00:05,333
through the night and no one
3
00:00:05,333 --> 00:00:08,000
in the house could sleep.
Three pieces of 2667, 2666 and 2667 ms — the rounding goes somewhere, and the last piece takes what is left so the span closes exactly on 8000 ms. The cuts fall after all and after one, which are the nearest word boundaries to a third and two thirds of the way through, so the pieces hold 25, 28 and 25 characters rather than 26 each.
A cue of exactly the limit, and one a millisecond over
Input
1
00:00:01,000 --> 00:00:04,000
Exactly three seconds long.
2
00:00:05,000 --> 00:00:08,001
Just one millisecond too long.
Output
1
00:00:01,000 --> 00:00:04,000
Exactly three seconds long.
2
00:00:05,000 --> 00:00:06,501
Just one millisecond
3
00:00:06,501 --> 00:00:08,001
too long.
The first cue is exactly 3000 ms and is left whole; the second is 3001 ms and is cut down the middle. That is the honest reading of "longer than the limit", and it is why a limit set one millisecond too low will double every cue that sits exactly on it. Check the durations in your file before you commit to a number.
A duration limit that reaches nothing
Input
1
00:00:01,000 --> 00:00:02,000
Short.
2
00:00:03,000 --> 00:00:03,900
Also short.
Output
1
00:00:01,000 --> 00:00:02,000
Short.
2
00:00:03,000 --> 00:00:03,900
Also short.
Thirty seconds is wider than anything here, so the file comes back untouched. Nothing is merged or stretched by a wide limit: this page divides cues and does nothing else, which makes it the safe way to find out what a file's longest cue actually is.
Frequently asked questions
Where does the seven-second limit come from, and is it the same everywhere?
Seven seconds is the most commonly quoted ceiling for one cue and it is a reading-speed rule in disguise: an average reader finishes roughly 15 to 20 characters a second, so a two-line cue of 84 characters needs somewhere near five seconds and a slow reader needs more. Broadcaster specifications differ — several European guides settle on six seconds — so treat the box as a starting point and check the specification you are delivering against. The tool applies whatever number is in the box, with no opinion of its own.
How does the tool decide how many pieces a cue becomes?
It divides the duration by your limit and rounds up, so 8000 ms against 3000 ms is three pieces and 8000 ms against 4000 ms is two. The pieces then take equal shares of the span, which is why the boundaries land on thirds and halves of the original cue rather than on the sentences or lines inside it. If you want the cuts to follow the text instead, the pages that split by characters or at sentence ends are the ones that do that.
Can I set a limit short enough to break a cue into a hundred pieces?
There is a ceiling of 200 pieces per cue, and a second one that matters more in practice: a piece cannot be shorter than a millisecond. So a nine-second cue with a limit of 1 ms gives you 200 pieces rather than 9000, and the report says the limit could not be reached. That is a guard against a stray keystroke turning one cue into a file of its own, not a limit you will meet with any realistic number.
Does dividing a long cue change the timeline of the whole file?
No. The pieces of a divided cue fill exactly the span the original cue occupied, and every other cue keeps its times to the millisecond. The only thing that changes elsewhere is the numbering, since a file whose cue count changed is written out numbered from 1 in timeline order. If a cue you divided was carrying a style, that style stays with the first piece and the report tells you how many new pieces came out unstyled.
My file has cues that linger because they hold one short line. Is this page the fix?
Only partly. This page can cut such a cue into two, but both pieces then hold a fraction of the same short line, which is not usually an improvement — the reader gets the same words in two helpings instead of one. A cue that lingers because nothing else is happening is better handled by fixing the timings themselves, which is what the shift and scale pages are for, and by checking whether the cue was meant to cover a pause in the dialogue.
Can a file be too long for this page to handle?
Files up to 64 MB are read, which is far more than any subtitle track: an hour of film is around 100 KB. Past that the page says so rather than failing quietly, because a file that size is almost certainly not a subtitle file. A very large real file — a full series, say — is better run through the queue one episode at a time, where each download can be checked before you move on.