What video formats does Sense support?

Sense supports common video formats such as .mp4 and .mkv by leveraging ffmpeg for high-quality frame extraction at specific timestamps.

Do I need existing subtitle files to use this skill?

Sense is designed to parse .vtt or .srt files. If they are not available, it can be integrated with tools like yt-dlp to generate auto-subtitles before processing.

Where is the extracted knowledge stored?

All extracted data, including timestamped transcripts and LaTeX equations, is stored in a local DuckDB database, allowing for fast SQL queries and custom view generation.

How does Sense handle mathematical equations in videos?

The skill integrates with the Mathpix API to perform specialized OCR on extracted video frames, converting both handwritten and printed mathematical notation into clean LaTeX code.

What is the purpose of GF(3) skill mapping?

GF(3) balance is a mathematical approach used by Sense to categorize content into a triadic structure (Validator, Coordinator, Generator), ensuring a balanced and structured representation of the knowledge extracted.

Sense - Video-to-Knowledge Extractor

Name: Sense - Video-to-Knowledge Extractor
Author: plurigrid

byplurigrid

•

学習とドキュメンテーション

Transforms video lectures into structured, queryable knowledge bases with subtitle alignment and LaTeX OCR.

概要

Sense is a specialized Claude Code skill designed to convert unstructured video content into structured, indexed knowledge. By combining subtitle parsing, frame extraction via ffmpeg, and advanced mathematical OCR via Mathpix, Sense extracts text, diagrams, and equations from lectures and presentations. It organizes this data into a local DuckDB database, providing a searchable timeline of technical concepts. The skill utilizes a unique GF(3)-balanced mapping system to categorize extracted content into specific domain skills, making it an essential tool for researchers and developers building technical documentation from video sources.

主な機能

Intelligent slide transition detection based on subtitle gaps and visual changes
Automated subtitle parsing and timestamp alignment for WebVTT and SRT files
High-fidelity diagram and equation OCR using Mathpix for LaTeX extraction
2 GitHub stars
Structured data storage in DuckDB for SQL-based knowledge retrieval
GF(3)-balanced skill mapping for domain-specific content categorization

ユースケース

Building automated knowledge graphs from online coding tutorials and workshops
Extracting technical diagrams and equations from video presentations for research
Indexing academic lecture series into searchable LaTeX-based documentation

概要

主な機能

Intelligent slide transition detection based on subtitle gaps and visual changes
Automated subtitle parsing and timestamp alignment for WebVTT and SRT files
High-fidelity diagram and equation OCR using Mathpix for LaTeX extraction
2 GitHub stars
Structured data storage in DuckDB for SQL-based knowledge retrieval
GF(3)-balanced skill mapping for domain-specific content categorization

ユースケース

Building automated knowledge graphs from online coding tutorials and workshops
Extracting technical diagrams and equations from video presentations for research
Indexing academic lecture series into searchable LaTeX-based documentation