youtube-clip-extractor_skill

Overview

This skill downloads YouTube videos, finds compelling clip moments from transcripts, cuts clips with ffmpeg, and produces platform-optimized on-screen text and captions. It delivers a complete URL-to-publishable-clips workflow so you can move from raw video to ready-to-post assets quickly.

How this skill works

It fetches the video and English subtitles using yt-dlp, analyzes the VTT transcript to identify 5–8 clip candidates that meet hook, arc, stakes, and coda criteria, and then extracts those clips with keyframe or frame-accurate ffmpeg commands. Finally it generates on-screen text hook options and platform-specific caption packages optimized for Instagram, TikTok, YouTube Shorts, Facebook, and LinkedIn.

When to use it

• You have a YouTube URL and want the best short-form clips extracted automatically
• You need clips with clear hooks and complete narrative arcs (30–90s)
• You want H.264 clips ready for editors like Descript
• You need on-screen text hooks and captions tailored to each platform
• You want a reproducible, folder-structured output for clip assets

Best practices

• Download H.264 (or re-encode) so clips import cleanly into editors like Descript
• Scan the full transcript and verify context before writing hooks or captions
• Prioritize clips with a strong first 3-second hook and a clear coda
• Use -c copy for fast cuts when keyframes align, re-encode for frame-accurate trimming
• Keep on-screen text to 2–4 words and match captions to the audio topic in the first sentence

Example use cases

• Extracting 6 short, publishable clips from a 45-minute interview for social distribution
• Turning a single long tutorial into platform-optimized micro-lessons with captions
• Batch-cutting highlight moments for a creator who needs quick upload-ready files
• Creating multiple hook variants and caption packages for A/B testing on TikTok and Instagram