Where is the metadata saved?

Metadata is saved in two places: a .meta.json sidecar file located next to the original media file and a centralized SQLite database file (media_catalog.db) at the root of the processed folder.

How does the AI analyze video files?

The skill extracts representative keyframes based on the clip duration and sends these visual samples to Claude to generate a comprehensive scene description and relevant tags.

Does this skill require external software?

Yes, it relies on ffmpeg for video processing and Python 3 for metadata handling. The skill will attempt to guide you through installation if they are not detected.

Can I search for specific visual elements?

Yes, you can query the catalog using the 'search' command to find assets by keyword, mood, location, or visual description (e.g., 'Find all clips of golden hour at the beach').

What file types does Media Ingest support?

It supports standard photo formats like JPG, PNG, and HEIC, professional RAW formats like CR2 and ARW, and major video containers including MP4, MOV, MKV, and BRAW.

Media Ingest & Semantic Cataloger

Name: Media Ingest & Semantic Cataloger
Author: michaeldboyd

bymichaeldboyd

0•

数据科学与机器学习

Automates the ingestion, visual analysis, and semantic tagging of photo and video assets into a searchable SQLite database.

Media Ingest transforms raw camera dumps into organized, searchable libraries by combining technical metadata extraction with AI-powered visual analysis. It automatically extracts keyframes from videos, generates rich natural language descriptions using Claude's vision capabilities, and assigns detailed semantic tags. Whether you are managing a professional footage archive or organizing a personal photo collection, this skill streamlines the workflow from SD card import to a fully indexed catalog with sidecar metadata and a centralized SQLite database.

主要功能

01Technical metadata extraction including EXIF, GPS, codecs, resolution, and camera settings.

020 GitHub stars

03Automated keyframe extraction from video files using ffmpeg at intelligent intervals.

04AI-powered visual analysis providing natural language descriptions and mood detection.

05Deep semantic tagging with 10-30 keywords per asset for high-precision searchability.

06Centralized SQLite cataloging with sidecar JSON files for portable metadata management.

使用场景

01Organizing and tagging large SD card dumps from professional cameras, drones, or mobile devices.

02Automating the generation of descriptive metadata for Media Asset Management (MAM) systems.

03Creating a searchable 'stock footage' style database of local media assets using natural language.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add michaeldboyd/media-ingest media-ingest

For use in Claude.ai and ChatGPT

Download Skill