Clean a YouTube transcript: remove timestamps, make paragraphs or an SRT

What it reads

When you select the whole transcript panel and copy it, the times usually come out on their own lines, with the words underneath. Some tools put the time at the start of the line instead. The cleaner reads both, and a mix of the two.

0:00
welcome back to the channel
0:04
today we are looking at how to clean up a transcript

0:09 so first of all open the video      (a time at the start of the line also works)

Lines that came along from the page, such as Show transcript or Search in video, and the spoken form of a time (1 minute, 5 seconds) are ignored and counted, so you can see what was dropped. A time in the middle of a line is not treated as a timestamp: “the call is at 10:30” stays as it is. Sound descriptions such as [Music] and [Applause] can be removed with one tick.

Three things you can get back

• Clean text in paragraphs. The timestamps are gone and the caption lines are joined into paragraphs that begin about every 15 seconds, 30 seconds, one minute or two minutes, whichever you choose. If the text you pasted has no times at all, paragraphs break after an estimated number of words instead, and the page says so.
• An SRT subtitle file. Each line becomes a cue that starts where YouTube says it did. Ends are calculated, long lines are wrapped to two lines of at most 42 characters, and a line that will not fit is split with its time shared out by length.
• A draft list of chapters. One chapter about every few minutes, starting at 0:00, each titled with the first words of the line where it starts. The result is checked against YouTube's rules for chapters (first time 0:00, at least three, ascending, ten seconds minimum) and the tool tells you if it fails. The titles are placeholders: it cannot know where the topic changes, so rename them.

What it cannot do

It does not add punctuation, fix misheard words, label speakers or find the real chapter breaks. It does not fetch a transcript from a link: YouTube's own panel is the source, and the quality of the text is whatever the captions were, automatic or uploaded by the creator. Treat an automatic transcript as a draft, read it against the video, and check names and numbers.

If you need to transcribe the sound itself

When a video has no captions, or the automatic ones are poor, the transcript has to come from the audio. If you have the file, upload it for transcription; the result has timestamps from the sound and exports TXT, SRT, VTT, Markdown, JSON and CSV. The page YouTube transcript explains the two routes, and remove timestamps and SRT to text cover other layouts. Nothing you paste here leaves your browser.