Cleaning video transcripts and turning them into text people read

What's inside a caption file

YouTube accepts several caption formats, including SubRip (.srt), WebVTT (.vtt) and SubViewer (.sbv). An .srt file is a list of numbered blocks: a counter, a start and end time such as 00:01:04,200 --> 00:01:07,900, then one or two short lines of text. WebVTT is similar, with a WEBVTT header and optional styling. YouTube's automatic captions often repeat each line in the next block so the words roll up on screen.

Copying from the transcript panel under a video gives a different mess: a timestamp on its own line before every few words, and sometimes spoken-time lines for screen readers.

What a cleaner removes, and what it can't fix

A transcript cleaner strips the counters, timestamps and formatting tags, removes sound descriptions such as [Music] if you want, drops lines repeated by rolling captions, and joins the rest into paragraphs. That turns a few hundred tiny blocks into readable text in seconds.

Reusing a transcript

Speech and writing work differently. Spoken text is full of repetition and filler that reads badly on a page, so treat the cleaned transcript as a first draft.

• Blog post or article: keep the structure of the video, add headings at each chapter, and cut repetition.
• Show notes: a short summary plus the chapter list and links mentioned in the video.
• Newsletter or social posts: pull out two or three quotable lines.
• Study notes or research: search a long talk for one passage instead of scrubbing through it.

Reading time versus speaking time

People read faster than they speak. A large 2019 review found that adults silently read non-fiction at about 238 words per minute, while a speaking pace is commonly put at around 150. A 10-minute video of about 1,500 spoken words becomes an article that takes six or seven minutes to read, before you cut anything.

Your own words are yours to reuse. Someone else's transcript is still their work: quote it briefly, with credit and a link.