Is this skill compatible with Windows or Linux?

Currently, this skill requires macOS as it depends on the Apple Vision framework and PyObjC for the OCR engine.

What output formats can I generate?

The skill supports JSONL for machine-readable data and Markdown for human-readable documentation, including the ability to embed the extracted frames as images.

How does the deduplication feature benefit my workflow?

It uses perceptual hashing to detect visually similar frames. This ensures that static parts of a video (like a single slide shown for several minutes) only get processed once, saving time and reducing output clutter.

Does this skill require an internet connection for OCR?

No, the OCR processing uses the native Apple Vision framework on macOS, meaning all text extraction and frame analysis happen locally on your machine.

Can I adjust the sensitivity of the frame comparison?

Yes, you can use the --dedupe-threshold argument to adjust how different two frames must be before they are considered unique.

Video OCR Toolkit

Name: Video OCR Toolkit
Author: varunr89

byvarunr89

0•

学習とドキュメンテーション

Extracts text from video files by sampling frames and performing high-performance Apple Vision OCR with perceptual deduplication.

The video-ocr skill provides a robust pipeline for converting video content into searchable text or structured documentation. It leverages ffmpeg to extract frames at a specific frequency and employs perceptual hashing (aHash) to intelligently discard visually similar frames, preventing redundant data processing. By utilizing the native Apple Vision OCR framework, it offers fast and accurate text recognition in parallel, making it an ideal tool for processing long recordings of presentations, tutorials, or lectures where on-screen information evolves over time.

主な機能

01Structured output in JSONL format for easy data ingestion

02Configurable frame extraction rate (FPS) via ffmpeg integration

030 GitHub stars

04Automatic Markdown generation with optional embedded frame snapshots

05Parallelized OCR processing powered by the native Apple Vision framework

06Intelligent frame deduplication using average perceptual hashing (aHash)

ユースケース

01Summarizing slide-deck presentations by capturing textual changes in on-screen content

02Extracting code snippets and text from technical software tutorials or screencasts

03Converting video lectures or webinars into searchable study guides and notes

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add varunr89/claude-marketplace video-ocr

For use in Claude.ai and ChatGPT

主な機能

01Structured output in JSONL format for easy data ingestion

02Configurable frame extraction rate (FPS) via ffmpeg integration

030 GitHub stars

04Automatic Markdown generation with optional embedded frame snapshots

05Parallelized OCR processing powered by the native Apple Vision framework

06Intelligent frame deduplication using average perceptual hashing (aHash)

ユースケース

01Summarizing slide-deck presentations by capturing textual changes in on-screen content

02Extracting code snippets and text from technical software tutorials or screencasts

03Converting video lectures or webinars into searchable study guides and notes

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add varunr89/claude-marketplace video-ocr

For use in Claude.ai and ChatGPT