From subtitle file to readable text
A subtitle file is mostly bookkeeping: a cue number, a time line like 00:01:02,500 --> 00:01:05,000, two short lines of text and a blank line, over and over. This tool keeps only the words, joins the short lines into sentences and paragraphs, and gives you text you can read, quote, translate or search.
It reads the four common formats. In SRT and SBV files the text is plain, in WebVTT and ASS files tags for italics, colors, speakers and positions are removed. WebVTT NOTE comments and STYLE blocks never end up in the text, and neither do ASS comment lines and vector drawings.
Repeated lines in rolling captions
Some caption files are written as rolling captions. A cue shows the line from before plus the new line, word timings sit inside the text like <00:00:01.500>, and a cue only 10 milliseconds long repeats the lines once more. Turned into text line by line, each sentence appears two or three times. The example in the tool is built that way, and the table shows what is kept:
| Cue | Text in the file | Kept |
|---|---|---|
| 1 | so today we | so today we |
| 2 (10 ms) | so today we | (repeat) |
| 3 | so today we / talk about subtitles. | talk about subtitles. |
| 4 (10 ms) | talk about subtitles. | (repeat) |
With Remove repeated lines on, a line is dropped when it says the same as the line right before it, ignoring case and punctuation. A line that only adds words to the line before it replaces that line, so text that grows word by word appears once, complete. Lines that repeat later on, like a chorus, stay. Switch the option off for scripts in which the same line really is said twice in a row.
Paragraphs, lines or one block
- Paragraphs start a new paragraph wherever the speech pauses for at least the time you choose. Two seconds suits most talks and interviews; lower it for fast dialogue, raise it for lectures.
- One line per cue keeps each subtitle on its own line, handy for checking a translation against the video.
- One block runs everything together, for pasting into a summary tool or a word counter.
Timestamps every few minutes mark where you are in the video. The stamp shows the start of the first line after each mark, so the times match what you see in the player. When a WebVTT voice tag or an ASS Name field gives a speaker, the name appears at the start of each new speaker's paragraph unless you remove speaker labels.
To count the words of the result, paste it into the Word Counter. To change the subtitle format instead, use the Subtitle Converter.
How to turn subtitles into text
- 1Paste the subtitles or drop an SRT, VTT, SBV or ASS file on the box.
- 2Pick a layout and, if you like, timestamps. Turn on the options that remove speaker labels or sound cues.
- 3Copy the text or download it as a .txt file.
Sources
- W3C, WebVTT: The Web Video Text Tracks Format: NOTE and STYLE blocks, voice tags and inline timestamps in cue text
- Library of Congress, SubRip Subtitle format (SRT): structure of an SRT cue
- Aegisub manual, ASS override tags: override blocks and line breaks in ASS text
Related guides
- How Many Pages Is 1,000 Words? Words per Page, Measured
- What Is a Token in AI? Tokens vs Words vs Characters
- How to Pick a Random Winner Fairly: Draws People Trust
Frequently asked questions
Why does every line appear two or three times in my transcript?
Your file is probably written as rolling captions: each new cue shows the previous line again above the new one, and very short cues repeat the text once more, so it scrolls up on screen. Turned into text as is, every line appears two or three times. Remove repeated lines keeps each line once, and a line that grows word by word is kept only in its full form.
How are paragraphs made?
A new paragraph starts where the speech pauses: when the next subtitle starts at least the chosen number of seconds after the previous one ends. Very long stretches without a pause are split after a sentence once a paragraph passes about 900 characters. One line per cue keeps every subtitle on its own line instead.
Can I keep timestamps in the transcript?
Yes. Pick an interval under Timestamps and the first paragraph after each mark starts with its time, like [5:02] or [1:15:40]. That makes it easy to jump back to a passage in the video. Choose Off for clean text.
What does Remove speaker labels take out?
Names at the start of a line followed by a colon, like JOHN: or Dr. Lee:, names in brackets like [Anna], the >> marks that broadcast captions use for a new speaker, and WebVTT voice tags. Be careful with text that starts lines with a word and a colon on purpose, such as Note:, since it looks the same.
Which files can I use?
SRT, WebVTT (.vtt), SBV and ASS or SSA files, pasted as text or dropped as a file. Formatting such as italics, colors and positions is removed, because a transcript is plain text.
Is my file uploaded?
No. The subtitles are read and turned into text by your browser on your own device. Nothing is sent to a server.