How to get a transcript of any YouTube video

Most of the time you do not want to watch the video. You want the words in it: to quote a line, to check what someone actually said, to turn a talk into notes, or to feed it to something else. There are four ways to get them, and which is worth your time depends almost entirely on whether the video has captions already.

1. YouTube's own transcript panel#

The fastest route, and the one most people never find. Under the video, open the ... menu and choose Show transcript. A panel opens with the caption text, timestamped; clicking a line seeks the player to it.

For reading along this is genuinely good and costs nothing. Its limits appear the moment you want to do something with the text:

• No search inside the panel — you scroll.
• No export. Selecting and copying is the only way out.
• Auto-generated captions carry no speaker labels and little punctuation, so a two-person conversation arrives as one undifferentiated column of words.
• If the video has no caption track, the option is not there at all.

2. Copy and paste into a document#

The manual version of an export: open the panel, select everything, paste, clean up.

For one video this is often the right answer. The cleanup is where it falls apart. Pasted transcript text carries a timestamp on its own line before every caption line, so a 15-minute video lands as several hundred alternating lines of number and fragment. Turning that into prose is a find-and-replace job you will do once cheerfully, twice grudgingly, and the third time you will go looking for a tool.

3. Get the caption file itself#

If you want a real subtitle file (.srt or .vtt) rather than text on a page, you need the caption track rather than the panel's rendering of it. This is what you want for editing video, re-uploading captions, or feeding a tool that expects timings.

Worth knowing, because it explains a lot of apparent inconsistency: YouTube's translated tracks are generated from an existing track. YouTube will happily offer a video's captions in a long list of languages, but only if the video already has at least one caption track to translate from. No original, no translations, however long the menu looks elsewhere.

It is also harder to do than it looks. There is no plain URL you can just call: YouTube's caption delivery is locked to the page itself, which is why browser extensions can read captions and a standalone script generally cannot.

4. Generate one with speech recognition#

When there is no caption track, the first three options do not exist. The only route left is to transcribe the audio.

In Tubelator this is the Transcribe tab. Two fields:

• Video Language — the language spoken in the video
• Target Language — translate subtitles into this language

Then Generate AI Subtitles and Transcript. The audio comes back as timestamped segments with per-word timings; when your target differs from the spoken language, a separate translation pass runs after it.

You will see the status change as it works: Accessing audio…, Processing video, then Transcribing… with a percentage, and finally Activating your new transcript. The percentage moves in visible jumps rather than sliding smoothly, because it tracks pieces of the job as they finish, not elapsed time. Text starts appearing within moments instead of after the whole file is processed, and segments stream into the Transcript tab while the job is still running.

That is also why long videos scale better than you would expect. Across nearly two thousand production runs the transcription stage had a median of about 40 seconds for a video of roughly 15 minutes, and around 80% of runs finished that stage inside a minute. A 90-minute video is not six times slower than a 15-minute one. Allowing for fetching and assembling the audio, "about a minute for a normal video" is the honest expectation.

Two things to know before you try it. Generating a transcript requires an account — reading captions that already exist does not. And a failed generation costs you nothing: usage is counted only when a transcript actually completes.

“

Its really good for other vids that don't have subtitles so i will be using this really crazy work!

— Shan Siva, Chrome Web Store review

Which one to use#

The honest summary: for a single video that already has captions, YouTube's panel and your clipboard are fine and anything else is overkill. The calculation changes when the video has no captions (where manual effort does not help at all, because there is nothing to copy) or when you are doing this often enough that the cleanup is the job.

Working with the transcript#

Once a transcript is loaded, the Transcript tab has a search field, an Export To button, an Analyze with AI shortcut, and a control to pause auto-scroll if you would rather it stopped following the video.

Two details worth knowing:

Search filters the view, not the export. If you search for a word and then export, you get the whole transcript, not just the matching lines. That sounds obvious and is a real bug in tools that get it wrong.

Every line has a ⋮ menu with Set Loop Start, Set Loop End, Copy and Speak. Looping is built for re-hearing a phrase rather than watching once and moving on: it repeats a passage until you clear it — see dual subtitles for language learning for the workflow this is built around.

Getting it out#

Export To opens Export Transcript. Two clipboard actions come first: Copy Full (with timestamps) and Copy Plain (no timestamps). Then four Popular Formats: .SRT (SubRip), .DOCX (Word), .PDF, .TXT. Show more formats reveals the rest.

One thing that is worth more than the count, because it is the sort of thing that fails quietly elsewhere: PDF export ships its own fonts. A PDF library's default font covers 256 characters, and anything beyond that comes out as mangled Latin or empty boxes without ever raising an error. Tubelator bundles Noto fonts covering Arabic, Chinese, Japanese, Korean, Devanagari, Thai, Hebrew, Tibetan and more, which also brings correct right-to-left ordering and contextual letter forms. If you have ever exported a Japanese transcript and got a page of rectangles, that is the bug being avoided.

Filenames get the same treatment. A / in a video title becomes a space rather than being deleted (dropping it welds Part 1/2 into Part 12, a different meaning rather than a missing character), and titles that collide with reserved Windows device names are renamed, because a file called CON.pdf simply fails to download.

For scale: the longest transcript measured so far, a 211-minute video, came to 1,876 segments.

What the free tier actually covers#

Unexpected paywalls are the most common complaint about tools in this category, so here are the real numbers instead of a "generous free tier" claim.

No account is needed to read captions that already exist, and free limits reset at the start of each calendar month.

Plus is $3.99 a month, 20% off the $4.99 list price, and the discount holds for as long as you stay subscribed. It raises every limit above and lifts the 30-minute video-length restriction. See the full pricing breakdown for Pro and the feature-by-feature comparison.

If you only ever need one transcript, use YouTube's panel and your clipboard. The tools are for the second, the fifth and the fiftieth.