Strip cue numbers, times and tags from SRT or WebVTT subtitles and get readable paragraphs. Nothing leaves your browser.
To turn an SRT or VTT file into text, remove the cue numbers, the time lines and the formatting tags, then join the remaining caption lines in reading order. This tool does that in one step. It reads each cue with the same parser as the SRT to VTT converter, so numbered SRT cues, WebVTT cues with ids, and cue settings such as align:start are all understood. The WEBVTT header, NOTE comments and STYLE blocks are dropped. Tags such as i, b, u and c are removed, along with the word timing tags that YouTube adds inside cues. Voice tags like v Anna are read as speaker names. HTML entities such as & and the ASS style blocks that some SRT files carry, like {\an8}, are cleaned too. A leading dash that marks a change of speaker in a cue is removed.
Subtitle cues are cut for the screen, not for reading, so a sentence is often split across two or three cues. In Paragraphs layout the tool joins cues with a space and starts a new paragraph when the pause between one cue's end and the next cue's start is longer than the number of seconds you set. The default is 2 seconds. A larger number gives longer paragraphs, and 0 starts a paragraph at every cue. One line per cue keeps the cue boundaries instead. With Keep speaker names on, a WebVTT voice tag becomes a Name: prefix and a change of speaker starts a new paragraph. SRT files have no standard way to name a speaker, so names appear only when the file uses voice tags. Time marker every N minutes adds a line such as [5:00] or [1:05:00] before the first text at or after each interval, which helps when you search a long talk.
Auto-generated captions, such as those downloaded from video sites, often roll up the screen: the second line of one cue becomes the first line of the next, so every sentence appears twice or three times in a plain text export. Remove repeated caption lines drops a line when it is the same as a line of the cue just before it, ignoring case and extra spaces, and counts how many it removed. It is off by default because it can also remove real repetition. If a speaker says no in one cue and no again in the next, with nothing else between them, the second one is removed. Only the cue directly before is compared, so a phrase that comes back later is kept. Check the result before you use it as a record of what was said.
The tool accepts up to 1,000,000 characters and 50,000 cues. A cue whose time line cannot be read is skipped. A file with no readable cue gets a message that explains what a cue needs. It works on text only: paste the contents of the file, because this page does not open files itself. The transcript does not include the time of each line unless you set the marker, and it does not fix spelling or add punctuation that the captions lack. To keep the timing but change the format, or to shift times, use the SRT to VTT converter. The text stays in your browser, in this tab. It is not uploaded or stored. Only your option choices are remembered.
Paste the text of the .srt file into the box. The cue numbers and times are removed and the lines are joined into paragraphs. Copy the result or press Download to save it as transcript.txt. A 3-cue file such as Hello, How are you?, New topic gives two paragraphs at the 2 second default.
Yes. Paste the WebVTT text. The WEBVTT header, the cue settings and the word timing tags inside cues are removed. For auto-captions that repeat lines, turn on Remove repeated caption lines to drop the lines each cue copies from the one before.
By the gap between cues. If the next cue starts more than N seconds after the previous one ends, a new paragraph starts. N is 2 seconds by default and can be 0 to 600. With 0, every cue is its own paragraph; with 600, nearly everything joins into one.
Yes, when the file has WebVTT voice tags such as v Anna. Turn on Keep speaker names and each paragraph starts with Anna: and a speaker change starts a new paragraph. SRT has no standard speaker field, so plain SRT files show no names.
It only compares a cue with the cue right before it, and it is off by default. Real repetition, like a speaker saying no twice in a row, is removed too. Read the result and turn the option off if a repeat was meant.
Yes. Set Time marker every to a number of minutes, for example 5. A line such as [5:00] is added before the first text at or after minute 5, and [1:05:00] after an hour. Set it to 0 for no markers.