youtubegptTranscript generator › Download a transcript
Download a YouTube transcript
Five file formats, each built for a different job. The trick is picking the right one first — converting between them afterwards is where people lose time.
Which format do you actually want?
Almost everyone arrives wanting "the transcript as a file" without a clear picture of what happens to it next. The next step is what decides the format, so start there:
| What you are doing next | Download this | Extension |
|---|---|---|
| Reading it, or pasting into a document | Readable | .txt |
| Quoting with references back to the video | Timestamps | .txt |
| Word counts, search, scripting over the text | Plain text | .txt |
| Feeding it to ChatGPT, Claude or Gemini | For AI | .md |
| Loading into a video editor or burning subtitles | SRT | .srt |
| Captions for an HTML5 player on the web | VTT | .vtt |
Downloading SRT when you wanted text is the common mistake
An SRT file opens fine in any text editor, which is why people grab it by default. But it carries a numbered block and a timecode line for every caption fragment, so a 3-minute video becomes about 4,100 characters of which roughly half is timing metadata. Stripping that back out by hand is tedious. If the destination is a document or a language model, take a text format instead.
The formats in detail
Readable — .txt
Caption fragments reassembled into paragraphs. Subtitles are timed in short bursts so they fit on screen, which means the raw stream arrives as hundreds of fragments cut at display boundaries rather than sentence boundaries. This format merges them back into paragraphs of roughly 700–1,200 characters, breaking at sentence ends where punctuation exists and at natural speech pauses where it does not. This is the one to take if a human is going to read it.
Timestamps — .txt
Every caption line prefixed with its timecode, one per line. Bulkier to read, but each line can be traced back to an exact moment in the video, which matters when you are quoting or fact-checking. The timestamps page goes into what the timecodes mean and how to jump to them.
Plain text — .txt
One continuous run of text, no timecodes and no line breaks. The right input for word counts, search, diffing, or anything that processes the text programmatically and would otherwise have to strip formatting first.
For AI — .md
Markdown with a short header carrying the video title, channel, duration and source URL, followed by the paragraphed body with a timestamp anchor at the start of each paragraph. Two reasons this beats pasting raw text into a chat window: the model gets the context it would otherwise guess at or ask you for, and the anchors give it something concrete to cite when it refers to a specific moment. It also survives the copy-paste round trip better than a wall of undifferentiated text.
SRT — .srt
SubRip, the most widely supported subtitle format there is. Numbered blocks, a start and end timecode using a comma before the milliseconds, then the caption text. Video editors, media players and most upload pipelines take it without complaint.
1
00:00:01,360 --> 00:00:03,040
[♪♪♪]
2
00:00:18,640 --> 00:00:22,000
♪ We're no strangers to love ♪VTT — .vtt
WebVTT, the format the HTML5 <track> element expects. Nearly identical to SRT with two differences that matter: a WEBVTT header line at the top, and a full stop instead of a comma before the milliseconds. Feeding an SRT to a web player that wants VTT usually fails silently — no error, just no captions — and that missing header is normally why.
WEBVTT
00:00:01.360 --> 00:00:03.040
[♪♪♪]How to download
- Paste the video link into the box above and press Get transcript.
- Choose the format tab that matches what you are doing next. Switching tabs does not re-fetch the video.
- Press Download. The file is saved with the video title as its name, so a folder of downloads stays identifiable.
If you would rather not download anything, Copy puts the current view straight on your clipboard. For pasting into a document or a chat window that is usually the faster path — see copying a transcript.
Notes and limits
File size
Transcripts are small. The longest video in our testing — a 266-minute course — produced 2,935 caption lines and 50,862 words, and even its SRT stays well inside a few hundred kilobytes. Length is not something you need to plan around.
Videos without captions
Nothing can be downloaded if the video has no caption track, because there is no text to begin with. This tool reads captions that already exist rather than transcribing audio itself. Most videos of any reach do have at least an auto-generated track.
Auto-generated tracks
When only an auto-generated track exists, the transcript is marked as such. It is genuinely useful but not proofread — speech recognition mishears names, jargon and anything spoken over music. Verify before you quote.
Copyright
Downloading a transcript does not transfer any rights to it. The words belong to whoever made the video. Studying, quoting and summarising are ordinary uses; republishing an entire transcript as your own content is not.
Frequently asked
Can I download transcripts in bulk?
Not from this page — it handles one video at a time. For batch work, call the API once per video ID from a script.
What is the difference between SRT and VTT?
VTT starts with a WEBVTT header line and uses a full stop before the milliseconds; SRT has numbered blocks and uses a comma. Use VTT for web players and SRT for editors and desktop players.
Can I download a transcript in a different language?
Yes, when the video carries a track in that language. Every available track is listed after extraction — see translated subtitles.
Does the downloaded file include the video title?
The file is named after the video. The For AI markdown format also puts the title, channel, duration and source URL inside the file; the other formats contain only the transcript.
More guides
Independent tool. youtubegpt.ai is not affiliated with, endorsed by, or operated by YouTube or Google. It works with publicly available caption data from YouTube and does not host, mirror or re-upload any video. Transcript text belongs to the original video creators. "YouTube" is a trademark of Google LLC.