Key Points
- Site converts long podcasts into readable bullet-point summaries instead of watching 3+ hour videos
- Must be run locally - no way to host publicly due to YouTube content restrictions and creator concerns
- Uses AI to process full YouTube transcripts and generate 18 bullet-point summaries
- Open source code available for local use to avoid legal issues with scaling
- Running at scale would be expensive due to high AI token costs for processing transcripts
- Amanda Knox example: 3-hour Joe Rogan podcast summarized effectively, revealing she's pursuing stand-up comedy
- Uses Firefox plugin called "Starrive" that started as YouTube video downloader using yt-dlp
- Backend runs on random port that Firefox plugin communicates with
- Tracks disk space usage as downloaded videos can quickly fill up storage
- Data folder contains all videos and can be deleted when space is needed
- Original version copied transcripts to clipboard for manual AI processing
- Current "Pod Papris" version automates the entire process without clipboard copying
- Works on any YouTube video by hovering and clicking when plugin is in correct mode
- Successfully tested on 10-hour long content with no issues
- Plugin shows "transcript downloading" and "generating bullets using OpenAI" status updates
- Simple workflow: hover over YouTube video, click, wait for AI processing, view summary
- Requires technical skills to set up - plugin installation, Firefox debugging, Go backend compilation
- Not suitable for non-technical users over 50 without computer skills
Full Transcript