sense - Extract indexed knowledge from lecture videos
Parses subtitles, extracts and OCRs key frames, maps content to skills, and stores timestamped results in DuckDB.
Tags
Updated: 2026-10-06Capabilities
Typical Inputs
Typical Outputs
What this skill does
- Parse VTT and SRT subtitles
- Detect slide timestamps
- Extract video frames
- OCR diagrams into LaTeX
- Map content to skills
- Balance skill trits
- Store indexed data in DuckDB
- Create timeline views
Inputs
- Video lecture file
- VTT or SRT subtitle file
- Output DuckDB path
Outputs
- Timestamped transcript records
- Extracted frame files
- OCR text and LaTeX
- Skill mappings with trits
- DuckDB index and timeline views
Requirements
- Ruby or Python runtime
- ffmpeg
- DuckDB support
- Subtitle parsing library
- Mathpix API credentials