YouTube Transcript

Where the words come from

Be clear about what this does before you rely on it. The tool does not listen to the video and it does not perform speech recognition. It asks our server to fetch the caption track that YouTube already holds for that video, and the server sends back the caption lines with their start times. Two kinds of caption track exist: ones the uploader wrote or uploaded themselves, and ones YouTube's own speech recognition produced automatically. Either will be returned. If a video has neither, because the uploader disabled captions or automatic captions were never generated for that language, there is nothing to fetch and you will see a message saying so. No amount of retrying will change that, and the fix is to transcribe the audio yourself.

Because a request goes to the server, this is not a browser-only tool. Only the video ID you paste is sent. The transcript itself is then held in the page, and the text, TXT and SRT files are all built locally from it. You pay nothing to use it. If your credit has been used up, a rewarded ad of a few seconds runs before the transcript is fetched.

Links the box will accept

The input is parsed rather than passed on blindly, so most things you copy out of a browser or a share sheet work:

• A full watch link with a ?v= parameter, including one that still has playlist or tracking parameters stuck on the end.
• A youtu.be short link from the Share button.
• A Shorts link, an /embed/ link, an old /v/ link, a /live/ link or a youtube-nocookie.com embed address.
• The bare video ID on its own, which is always exactly 11 characters made up of letters, digits, underscores and hyphens.

Anything the parser cannot reduce to a valid 11 character ID is refused before any request is made. Channel pages, playlists with no video in them and search result addresses fall into that group.

Reading it, and the three ways out

Each caption line becomes a row with its start time on the left as a cyan link. Times under an hour are shown as minutes and seconds, and anything past the hour mark gains an hours field. Clicking a time opens that video in a new tab at that second. The header line shows the video title, the caption language code when the server reports one, and how many segments came back, which is a quick check that you got the whole thing.

The Show timestamps checkbox does double duty. It hides the times in the reading view, and it also decides whether the Copy text button and the TXT download include them. Turn it off for clean prose to paste into notes, leave it on when you need to cite a moment. The SRT download ignores the checkbox and always writes proper subtitle timing: cues numbered from one, each running from its own start until the next segment begins, in the HH:MM:SS,mmm format that VLC, Premiere Pro, DaVinci Resolve and most players expect. The final cue has no following segment to end against, so it is given a four second duration. Both downloads are named after the video title, with punctuation stripped and the name trimmed to 60 characters, falling back to transcript when the title is unusable.

A worked example

Paste https://youtu.be/aBcD1234_-x and the parser pulls out aBcD1234_-x as the ID. Suppose the caption line "and then we add the flour" starts at 92.5 seconds. In the reading view it appears as 01:32, linking to youtube.com/watch?v=aBcD1234_-x&t=92s. With timestamps switched on, the copied text and the TXT file both read [01:32] and then we add the flour. In the SRT, if the following line starts at 95.2 seconds, that cue is written as 00:01:32,500 --> 00:01:35,200. A video titled "How to Bake: Sourdough (Part 2)" downloads as How to Bake Sourdough Part 2.txt.

What people do with it

Students turn a two hour lecture into searchable notes and jump back to the bit that made no sense. Journalists and researchers pull exact quotes from interviews without scrubbing a timeline. Creators repurpose their own videos into blog posts and show notes. Anyone who reads faster than people speak can decide in half a minute whether a long tutorial covers their problem. It also gives a text version to someone who cannot use audio, and helps language learners read along with a native speaker.

Accuracy and copyright

Automatic captions are good, not perfect. Names, technical jargon, overlapping speakers, accents and background music all cost accuracy, and automatic tracks carry no punctuation conventions you can depend on, so proofread anything you intend to publish. There is also no speaker labelling, because the caption format does not carry it. On rights, a transcript is still the creator's words. Quoting a passage with a link and attribution is ordinary practice; republishing an entire transcript as your own content is not.

When to use a different tool

For your own recordings, lectures you taped or a podcast file on your drive, use Audio to Text Transcription, which works from an uploaded file instead of a caption track. Pull the audio out of a video first with MP4 to MP3. To style captions for a social clip, see Caption Studio. Once you have the text, the Word Counter tells you how long the talk actually was in words, the Readability Checker rates a transcript you are turning into an article, and the Citation Generator builds a reference for the video you are quoting.

Questions

Which YouTube link formats work?

Standard watch links, youtu.be short links, Shorts, embed, live and old /v/ addresses, plus youtube-nocookie embeds. You can also paste the bare 11 character video ID. Extra playlist or tracking parameters on the end are ignored.

Why does it say the video has no captions?

Because the tool can only return a caption track that already exists on YouTube, either written by the uploader or generated automatically. If the uploader turned captions off and none were auto-generated, there is nothing to fetch, and the tool cannot transcribe the audio itself.

Does the video get uploaded or downloaded?

Neither. Only the video ID is sent to our server, which retrieves the caption data for it. The video file is never touched, and the text, TXT and SRT are assembled in your browser from what comes back.

How are the SRT cue times worked out?

Each cue starts at its caption line's own start time and ends where the next line begins, so the subtitles run continuously with no gaps. The last cue has nothing after it, so it is given a four second duration.

Can I get a transcript of a private or members-only video?

No. The caption data has to be publicly retrievable, so private, unlisted with restricted access, age-gated and members-only videos will return nothing.