Original video in Chinese.
Key Takeaways
- Obsidian Web Clipper saves articles and full YouTube transcripts as local Markdown in your chosen vault location, making them accessible to AI tools such as Claude Code.
- Its Interpreter can preprocess content with a model such as Gemini. Templates and triggers select the appropriate workflow automatically.
- I combine it with newtype OS for a weekly review, connecting collection, extraction, and deeper analysis.
If you use Claude Code with Obsidian, I strongly recommend Obsidian Web Clipper. It can substantially improve the efficiency of collecting information.
The extension extracts webpage content in the format you specify and saves it to your chosen location in Obsidian.
For AI users, the valuable part is that each saved item is a local Markdown document. Almost any AI tool can read it easily and process it further. If you use a tool such as Claude Code, you will appreciate that.
Preprocess information with the Interpreter
The extension has its own AI capability. Enter an API key, model ID, and name in Interpreter to let a model preprocess information.
I used Gemini 3.1 Flash through OpenRouter because it was inexpensive, had a large context window, and was sufficient for summarization and extraction.
I introduced this extension in a video the previous year. I recommended it again because a recent update could extract full YouTube transcripts by default.
In the demonstration, I made no special settings and disabled the AI feature. I opened a YouTube video and clicked the extension button. After a second or two, it had extracted the transcript.
Previously, this took more configuration and interaction. You had to open YouTube’s transcript display before using the extension; otherwise, it could not extract it.
My current workflow extracts the transcript and summarizes it so I can decide whether the video merits deeper attention.
You could also ask it to translate and tidy the full transcript into Chinese. English is not an obstacle for me, so I skip that step.
Two templates and a weekly review
I use two templates. One clips articles, mostly WeChat public-account articles, and asks Gemini 3.1 Flash to summarize and extract the main information.
The other clips YouTube videos. I added a Template Trigger so that a YouTube URL automatically selects it.
Every weekend, I use newtype OS to process everything saved during the week. I created a custom command for deeper analysis that tells me what deserves attention and why.
The extension is simple to use, but it has helped me considerably. If you regularly collect and digest large amounts of information, try it.
That is all for this episode. If you want to understand AI, become a more capable independent creator, and meet people with similar interests, join the newtype community. See you next time.