Split Subtitles at Sentences
Or paste subtitle text
Options
Cuts wherever one sentence ends and the next begins, and also wherever the cue changes line, because subtitle files often leave the full stop out. A full stop, question mark or exclamation mark that starts a new sentence counts, including the full-width 。！？; a lower-case continuation such as e.g. this, a decimal such as 3.5 and a common abbreviation such as Mr. or Dr. do not. Each sentence then gets a share of the cue's time in proportion to its text.
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 puts one sentence per cue. A transcription export often writes several sentences into a single cue — one cue per speaker turn, or per automatic paragraph break — and this page divides each of them where one sentence ends and the next begins, then shares the cue's time between them in proportion to how much text each sentence carries.
There is no number to fill in. The break points come from the text: a full stop, a question mark or an exclamation mark that starts a new sentence, including the full-width versions used in Chinese and Japanese. A break also falls wherever the cue changes line, because subtitle files are frequently written one line per sentence with the stops left out — without that rule a file like that would look like one the tool refuses to touch.
The judgement about what ends a sentence is a heuristic and it leans towards leaving cues alone. A lower-case continuation such as e.g. this, a decimal such as 3.5, and common abbreviations such as Mr., Dr. or etc. do not end a sentence; a dot between two digits never does. Miss an abbreviation and the worst case is one cue the tool did not divide, which you can see and fix by hand. Be eager about it instead and you cut names in half, which nobody can undo by reading the output.
Everything runs inside this browser tab — no file is uploaded. SRT, WebVTT, ASS, SSA, SBV, SAMI, MicroDVD, TTML, JSON and CSV are read the same way, and a file whose cue count changed is written out in timeline order numbered from 1.
Examples
Three sentences in one cue
Input
1
00:00:01,000 --> 00:00:04,000
He left. She stayed. We waited.
Output
1
00:00:01,000 --> 00:00:01,828
He left.
2
00:00:01,828 --> 00:00:02,966
She stayed.
3
00:00:02,966 --> 00:00:04,000
We waited.
The three seconds are shared 828, 1138 and 1034 ms — in proportion to 8, 11 and 10 characters. Reading speed is the point of the split, so the longest sentence keeping the screen longest is the intended outcome and not a side effect of the arithmetic.
Dots that do not end a sentence
Input
1
00:00:01,000 --> 00:00:03,000
Mr. Smith left at 3.5 seconds.
2
00:00:05,000 --> 00:00:08,000
Dr. Jones arrived. The room was full.
Output
1
00:00:01,000 --> 00:00:03,000
Mr. Smith left at 3.5 seconds.
2
00:00:05,000 --> 00:00:06,500
Dr. Jones arrived.
3
00:00:06,500 --> 00:00:08,000
The room was full.
The first cue holds three dots and none of them ends a sentence, so it comes back whole. The second holds two real sentence ends and becomes two cues, with Dr. left alone again. This pair is the whole judgement in one file: what the abbreviation list covers decides what happens here.
One sentence per line, with no stops at all
Input
1
00:00:01,000 --> 00:00:04,000
He said hello
then she answered quietly.
Output
1
00:00:01,000 --> 00:00:02,000
He said hello
2
00:00:02,000 --> 00:00:04,000
then she answered quietly.
There is no full stop between the two lines, so the line break itself is treated as the end of a sentence. Files written that way are common in transcription exports, and this rule is what makes them dividable at all. Each piece gets half the span here because the second line's 26 characters are twice the first line's 13.
Frequently asked questions
How does this page know where one sentence ends and the next begins?
It looks for a full stop, question mark or exclamation mark that starts a new sentence, plus the full-width versions used in Chinese and Japanese, and it treats every line break inside a cue as a sentence end as well. Three things stop it counting a dot: a lower-case letter following, as in e.g. this; a digit on both sides, as in 3.5; and a word on its own abbreviation list, which holds Mr., Mrs., Dr., Prof., St., Jr., Sr., vs., etc., fig., no., vol., approx., inc., ltd., co., dept., a.m., p.m., u.s. and u.k., along with any single letter such as J.
A cue holds a sentence my file wrote without a full stop. Will it be divided?
Only if the cue changes line there. A line break counts as a sentence end, so a file written one sentence per line divides even with no punctuation at all, which is how most transcription exports look. A file that runs two sentences onto one line with no stop between them gives the tool nothing to work with, and that cue comes back whole — the report says how many cues could not be divided rather than pretending the run did something.
Is splitting at sentence ends the same as splitting at punctuation marks?
No, and the difference matters with quotations and brackets. A closing quote or bracket after the stop stays with the sentence that ends — He said "Stop." And left. divides after the closing quote, not between the full stop and the quote. A dot inside a number, an initial or an abbreviation is not a sentence end at all, so a cue like Dr. Jones arrived at 3.5 p.m. stays whole until the next real stop or line break.
Do the divided sentences keep the times they had in the original cue?
They cannot — the original file only recorded when the whole cue was on screen, and nothing in it says where inside that span one sentence gave way to the next. So the pieces take shares of the span in proportion to their character counts, which stands in for reading speed: a sentence with twice the text stays up roughly twice as long, and the first piece starts and the last ends exactly where the original cue did.
Can I use this page to turn a transcript into subtitles one sentence at a time?
Yes, if the transcript is already a subtitle file with timings — the page works on cues, not on prose, so a plain text file has no cue boundaries to divide and no timings to share. What it will not do is invent the timings for a file that never had any. Bring the file in as SRT or convert it from plain text first, then divide it here; the pieces will be one sentence each, ready for a pass on the timings.
What happens to the rest of a cue's formatting when it is divided this way?
Style, position and karaoke timing belong to the cue they were written on, so the first piece keeps them and the pieces created after it have none. That is visible on styled formats such as ASS and on WebVTT cues with positioning, and the report counts how many new cues came out without formatting. For a file with heavy styling per cue, dividing by hand in a subtitle editor may be less disruptive than running the whole file through.