Video Scraping
Skill Seekers v3.2.0 introduced a complete video tutorial extraction pipeline that converts YouTube videos and local video files into AI-consumable skills.
Quick Start
# YouTube video
skill-seekers video --url https://www.youtube.com/watch?v=... --name mytutorial
# Local video file
skill-seekers video --video-file tutorial.mp4 --name mytutorial
# YouTube playlist
skill-seekers video --playlist https://www.youtube.com/playlist?list=... --name series
# First time setup (installs GPU-aware deps)
skill-seekers video --setup
What Gets Extracted
The video pipeline extracts multiple data streams:
- Metadata — Title, channel, views, chapters, duration (via yt-dlp)
- Transcripts — Multi-source with 3-tier fallback: YouTube Transcript API → yt-dlp subtitles → faster-whisper local transcription
- Segments — Chapter-based or time-window segmentation with configurable overlap
- Visual code — OCR on code editor panels, code timeline tracking across frames
- SKILL.md — Structured output combining all extracted knowledge
Optional Dependencies
pip install "skill-seekers[video]"
This installs: yt-dlp, faster-whisper, pytesseract, opencv-python, and other visual extraction dependencies.
Pipeline Architecture
The video pipeline follows this flow:
Video → Metadata → Transcript → Segmentation → Visual Extraction → SKILL.md
Key modules:
video_scraper.py— Main orchestrator (~960 lines)video_metadata.py— YouTube/local metadata extractionvideo_transcript.py— Multi-source transcript with fallbackvideo_segmenter.py— Chapter-based and time-window segmentationvideo_visual.py— Visual extraction with OCR (~2,410 lines)
Also Available via Create
# Auto-detection routes YouTube URLs to video scraper
skill-seekers create https://www.youtube.com/watch?v=...