openwiki: turn YouTube videos, blogs and notes into a Markdown knowledge base

View on GitHub

OpenWiki turns YouTube videos, blogs and your own notes into a linked Markdown wiki you keep. It's an open-source command-line tool: it downloads transcripts and articles, then an LLM writes a page for every source, person, company and topic, all cross-linked and ready for Obsidian.

source on github, get started, or sponsor the project.

sources

YouTubevideos · playlists · channels Blogs & essaysany site Your notes.md · .txt · .html

the wiki

Linked pagessources · people · topics Obsidian vault[[wikilinks]] Any LLMgemini · openai · ollama

get started

Python 3.10 or newer. Grab a video and an essay without any keys, then add an LLM to build the wiki.

$ pip install "git+https://github.com/ckryptickunal/OpenWiki.git"
$ openwiki init --root my-wiki && cd my-wiki
$ openwiki youtube https://youtu.be/jNQXAC9IVRw --folder Talks
$ openwiki ingest --all
  • free & MIT
  • no key for single videos
  • runs offline with Ollama
  • incremental

questions

How do I download the transcript of a YouTube video as text?

Run openwiki youtube <url> --folder Talks. It writes a .txt file with the title, channel, language and full transcript. No API key is needed.

How do I turn a whole YouTube channel or playlist into notes?

Set a free YOUTUBE_API_KEY, run openwiki youtube --channel @handle or --playlist <url>, then openwiki ingest --all to build the wiki pages.

Does it work with non-English videos?

Yes. Use --lang to prefer languages; if none match, OpenWiki uses whatever caption track exists. Page names keep non-Latin scripts.

Can I use it with Obsidian?

Yes. Open the wiki folder as a vault. Pages use YAML frontmatter and [[wikilinks]], so backlinks and the graph view work.

Can I run it without sending data to the cloud?

Yes. Point it at Ollama or LM Studio. Only the caption and article downloads touch the internet.

What does it cost?

OpenWiki is free and MIT-licensed. You only pay for LLM calls if your provider charges; a local model costs nothing.