YouTube transcript → Markdown

How to export a YouTube transcript as Markdown

There are two clicks between a YouTube video and a Markdown file you can paste into ChatGPT or drop into Obsidian. Replace youtube.com with ytscribe.io in the URL bar; the transcript loads. Open the dropdown next to the Copy button and choose Copy as Markdown. That's it. The Markdown is in your clipboard.

What you get back is structured: a level-1 heading with the video title, a small front-matter block carrying the source URL, language, and snippet count, then the transcript itself as proper paragraphs. The full timestamped view is appended below the paragraphs so you can keep timing data without losing readability.

1. Open the YouTube video in your browser.
2. In the URL bar, change youtube.com to ytscribe.io. Press Enter.
3. Click the dropdown arrow next to Copy and choose Copy as Markdown.
4. Paste into ChatGPT, Claude, Notion, Obsidian, or any tool that understands Markdown.

Why Markdown beats plain text for transcripts

Plain text loses structure. When you ask an LLM to summarise a 90-minute talk, the model has no signal that there's a title, who said it, or what the source was — it has to infer all of that from the transcript itself, which dilutes attention. Markdown front-matter solves this in two lines: the title gets cited correctly, the source URL becomes a clickable anchor in the response, and the language tag helps the model handle non-English transcripts without translation drift.

Markdown also survives copy-paste between tools. Notion, Obsidian, Logseq, and Roam all parse it natively. ChatGPT and Claude render it with proper headings and code blocks. A plain-text transcript, by contrast, loses the title-vs-body distinction the moment it lands in Notion's editor.

Example — Andrej Karpathy's LLM intro

Take Andrej Karpathy's "Intro to Large Language Models" — a one-hour talk that's frequently shared in dev circles. Open it on YouTube, swap the domain, click Copy as Markdown, paste into ChatGPT with a prompt like "Summarise this talk in five bullet points and three follow-up questions." You get a useful answer in about ten seconds. Without Markdown front-matter, the same prompt produces a generic summary that doesn't reference the speaker or the talk title.

You can try it now on the Karpathy talk if you want to see the exact Markdown output before committing to the workflow.

FAQ

Does the Markdown include timestamps?

Yes. The Markdown export includes the timestamped view as a second section after the paragraph view. If you only want timestamps, switch to the Timestamped tab and use plain Copy — that copies just the timestamped lines.

Can I paste this directly into Obsidian or Notion?

Yes. Both tools parse Markdown natively. Pasted ytscribe output renders with the title as a heading, the front-matter as a quoted block, and the transcript as proper paragraphs — no manual formatting required.

How long can a transcript be?

There is no hard cap. ytscribe has handled multi-hour podcasts and lecture courses without trouble. The practical limit is whatever your downstream tool accepts — ChatGPT and Claude handle hundreds of thousands of tokens in their long-context modes, Obsidian and Notion have no length limits.

Will Copy as Markdown work on YouTube Shorts?

Yes. Shorts under ytscribe.io/shorts/VIDEO_ID render the same way as regular videos and the Markdown export is identical. If a Short doesn't have captions, you'll see a clear message rather than empty Markdown.