video-search

Search recorded video by natural language queries against semantic embeddings and CV attributes.

Updated May 23, 2026
One-click install
npx skills add https://github.com/yo-steven/skills-exploration-20260522 --skill video-search-yo-steven
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: video-search
Source: https://github.com/yo-steven/skills-exploration-20260522/tree/main/skills/video-search-and-summarization/video-search
Command: npx skills add https://github.com/yo-steven/skills-exploration-20260522 --skill video-search-yo-steven

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill eliminates the time-consuming process of manually scrubbing hours of recorded footage by letting you search video archives with natural language and get timestamped results.

Core Features & Use Cases

  • Natural-language video archive search: Locate events, objects, actions, and people using semantic (embedding) matching.
  • Fusion search for higher accuracy: Combines Cosmos Embed1 semantic similarity with CV attribute search when the query includes both actions and appearance attributes.
  • Clip-focused results for investigation: Returns ranked matches with timestamps and playback/screenshot links suitable for verification.

Quick Start

Use the video-search skill with the deployed VSS search profile to search ingested videos by asking for the specific object, action, person, or time window you want to find.

Frequently Asked Questions about video-search

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I search recorded video for specific events using natural language?

Video search using natural language matches your queries against stored semantic embeddings and computer vision attributes to return timestamped clips of events, objects, actions, and people.

Can I search ingested video for a vehicle or staff activity within a specific time window?

Targeted video search supports locating specific vehicles or staff activity within given cameras and time windows by matching natural-language queries to stored semantic embeddings and CV attributes.

How does semantic video search combine appearance attributes with actions?

Fusion video search combines Cosmos Embed1 semantic similarity with computer vision attribute search when your query includes both actions and appearance attributes, yielding higher accuracy results.

Do I need an active VSS search profile to find timestamped clips in video archives?

Yes, video search requires an active VSS search profile on the host at the specified IP address with REST generation enabled to locate timestamped clips in your video archives.

Can I verify video search results using screenshots from the returned hits?

Video search results include timestamped playback links and an optional screenshot_url for results verification, allowing you to visually confirm the matched events, objects, or people.