C
ChaoBro

qiaomu: A Claude Skill That Turns Any Content Into NotebookLM Material—Podcasts, PPTs, Mind Maps

qiaomu: A Claude Skill That Turns Any Content Into NotebookLM Material—Podcasts, PPTs, Mind Maps

Google's NotebookLM is good. Upload documents, it generates podcasts, does deep Q&A, produces summaries. But it has a致命 limitation: you can only manually upload files.

WeChat articles? No. YouTube videos? No. Deep reporting behind paywalls? Even less so.

qiaomu (originally named anything-to-notebooklm) exists to tear down this wall.

What It Does

A Claude Code Skill that lets you tell the AI in plain language: "Turn this WeChat article into a podcast." "Make this YouTube video into a PPT." "Convert this podcast episode into a mind map."

AI auto-completes: content fetching → NotebookLM upload → target format generation. You just say one sentence in human language.

Content Source Coverage

The coverage range is wider than expected:

  • WeChat Official Accounts (MCP browser simulation scraping)
  • X/Twitter (tweets + full threads)
  • YouTube (auto subtitle extraction)
  • Podcasts (Xiaoyuzhou, Ximalaya, Bilibili)
  • Paywalled news sites (300+, including NYT, WSJ, FT, Bloomberg, The Information)
  • Any public webpage
  • PDF (including scanned OCR), EPUB, Markdown, Word, Excel
  • Images (auto OCR), audio (auto transcription)

Paywall Bypass

The狠 part of this project is the built-in paywall bypass. 6-level cascade strategy, from light to heavy:

  1. Proxy services (r.jina.ai / defuddle)
  2. Site-specific Bot UA (Googlebot covers ~50 sites, Bingbot ~4)
  3. Generic bypass (UA spoofing + X-Forwarded-For + Referer + AMP + EU IP)
  4. archive.today archive
  5. Google Cache
  6. agent-fetch local tool

Technical route borrowed from BPC (Bypass Paywalls Clean). NYT, WSJ, FT, The Economist, The Information... basically all tier-1 paywalled media are covered.

This isn't a small feature. For people who need to read大量 paywalled content, this capability alone is a product.

My Take

This project's value is in reducing the friction between content acquisition and knowledge processing to near zero.

Previously, turning a paywalled deep report into a podcast required: find a bypass method → copy content → manually upload to NotebookLM → wait. Now you just say one sentence to Claude Code.

A few notes:

  • Paywall bypass exists in a legal and ethical gray area. The tool is open-source, but judge the use case yourself.
  • NotebookLM is a Google closed-source service; you can't control the privacy policy after content upload.
  • The project is young (11 commits, 4 months); stability and long-term maintenance need observation.
  • Podcast transcription depends on a third-party API (Get笔记); there are cost and availability risks.

If you're a heavy NotebookLM user who frequently needs content from non-standard sources, qiaomu is currently the smoothest bridge. If you only occasionally use NotebookLM for local files, manual upload is sufficient.

2,347 stars in a week shows the demand is real. Worth watching.

Main sources: