Skip to content

Web Clipper and YouTube

The Obsidian Web Clipper is a browser extension that saves what you see as a Markdown file. It is the way to get in pages behind a login, LinkedIn, X and Reddit posts, and YouTube transcripts: the worker never scrapes those, and the clipped text is the source. It is optional, and it needs Obsidian.

Set it up with the wizard

sb setup clipper

The wizard has five stages:

  1. Install the extension for your browser and pin it to the toolbar.
  2. Add your vault to the Clipper's settings, by the vault's folder name.
  3. Import the template clipper-template.json. It saves into inbox/ with the properties title, source (the page URL), author, published and kind: article. The wizard also tells you about the YouTube template (below).
  4. Clip a test page (the wizard opens one).
  5. Run sb scan to check the worker reads the clip.

If your config says viewer = "none" the wizard says so and stops: the Clipper saves into an Obsidian vault.

Use it

Clip a page with the extension. The next sb run, or tonight's run, turns it into a note. The clip is not fetched again: whatever you saw is what is read, which is why this works for pages that need a login.

To queue clips without running everything: sb scan.

What is removed from a clip

Clips of GitHub, LinkedIn, X/Twitter, Reddit and YouTube pages come with the site's menus, sign-in prompts, cookie lines, sidebars, share bars and footers. Before the model reads the clip, the worker removes the lines that are known page chrome, so the note summarises the post or article and not the menu. Only whole lines that are known chrome are removed (never a sentence that contains such words, never code, plain bullets or transcript rows), and a clip from any other site is not touched. The original stays in raw/inbox/; the note's stripped_lines: N and the run's warning tell you how many lines were taken out. A clip that was nothing but chrome is rejected as having almost no text. Details: Add a Markdown clip.

YouTube videos with a transcript

Import the second template, clipper-youtube-template.json. It applies by itself on youtube.com/watch pages. Open a video that has a transcript and clip it.

The note is written like any other, from the transcript, and every quote and glossary term that the transcript contains carries a link that opens the video at that second: ([2:20](https://youtube.com/watch?v=...&t=140s)). The transcript is whatever language the Clipper gets from YouTube (often English, even for a Spanish video); the note is written in the language of [notes].language. A video without a transcript cannot be processed this way. The worker never downloads video or audio.

Without the Clipper

Any tool that saves a page as a Markdown file with title: and source: (the URL) at the top works the same if you drop its files in inbox/. See Add a Markdown clip.