Transcript-first filtering
Each suggested video is identified by URL/ID, transcript text is fetched, and only then scored with the model configured in your backend.
Skiper
Signal over noise for YouTube
Skiper blocks low-value recommendations before they steal your time.
Skiper reads the spoken content of recommended videos, scores quality and personal relevance with your own preference prompt, then masks weak recommendations while keeping everything clickable.
Personal relevance engine
Your custom prompt is always included in scoring. Relevance is measured against what is actually useful for your goals.
Live preference evolution
Use the Feedback button on the current video: your comment and transcript update your personal prompt, so the feed learns your standards.
How to install and run Skiper
Full setup in a few minutes. This build is distributed as a ZIP because it is not yet in the Chrome Web Store.
Click "Download Extension" above and save the archive to your machine.
Extract the ZIP into any folder, for example `~/Downloads/skiper-extension`.
Open `chrome://extensions`, enable Developer Mode, click Load unpacked, and choose the extracted folder.
Open Skiper from the toolbar and write what content is useful for you (topics, depth, style, what to avoid).
Open Home, Subscriptions, or any watch page. Skiper will score recommendations and hide videos with score below 5.
On any watched video click Feedback near the title, explain what was good/bad, and Skiper updates your prompt automatically.