Text output
YouTube to text
Get the words out of a video and into something you can edit. Copy with timestamps for reference, or without them for prose you can paste straight into a draft.
Works with public videos that have captions. Free — no sign-up.
Four things people do with the text
Turn a lecture into study notes
Copy without timestamps, paste into your notes app, then cut everything that is not a definition or an example. A fifty-minute lecture usually collapses to two pages. The timestamps in the original tab are still there if you need to re-watch a proof.
Quote a podcast in a piece of writing
Search the transcript for the phrase you half-remember, click the timestamp to hear it said, then copy the line. You end up with the exact wording and the exact moment, which is what a citation needs.
Feed it to a language model
Paste the plain text into ChatGPT, Claude or a local model and ask for a summary, an outline or the three claims worth checking. Markdown keeps the timestamps as bold markers, so the model can cite a moment rather than a vague section.
Find the one part you actually need
A two-hour interview has about four minutes you care about. Search for the term, read the surrounding lines, click through to the moment. That is faster than scrubbing the timeline and hoping.
Which copy format to pick
| Format | Use it when |
|---|---|
| Plain text with times | You will come back to the video and need to find moments. |
| Plain text, no times | The words are going into a document, an email or a model. |
| Markdown | You want a titled, formatted note with the source link already in it. |
| JSON | You are writing code against it — every cue with a start and end in seconds. |
Working with the text
Can I get the text without timestamps?
Yes — that is the second item in the Copy menu. You get continuous prose with one line per caption cue and nothing else, which is what you want when the text is going into a document rather than a reference.
Why does the text read in short fragments?
Because captions are written to fit on screen for a second or two, not to be read as prose. Sentences break mid-clause and punctuation is sparse in automatic captions. Expect to spend a couple of minutes joining lines and adding full stops if the text is going into something you will publish.
Is it accurate enough to quote?
Treat a quote as a draft until you check it. Click the timestamp next to the line, listen to those few seconds, and confirm the wording. That takes about ten seconds per quote and catches the misheard names and numbers automatic captions are prone to.
Can I paste this into ChatGPT or Claude to summarise it?
Yes, and it is one of the most common uses. Copy without timestamps for the cleanest input, or copy as Markdown if you want the model to be able to point back to a specific moment. A one-hour talk usually lands well inside a modern model's context window.
Does copying send anything to your server?
No. Once the transcript is on your screen, every copy and every download is produced in your browser from data you already have. There is no second request.
Get the words out of a video
Paste a link, then use the Copy menu to take the text in whichever shape your next tool wants.
Get a transcript