skills/modelscope.cn/watch-video

watch-video

SKILL.md

Watch Video

Analyze a public video URL and produce a vivid, detailed report of what was seen on screen AND said aloud — precise enough for an engineer, PM, or designer to act on without watching it.

Workflow

Step 1: Determine video platform

Platform Strategy
Loom Extract metadata + thumbnail from page; use Gemini on thumbnail + description
YouTube Use yt-dlp to download audio → Whisper transcription; or Gemini with thumbnail
Other Try yt-dlp first; fall back to page scrape + thumbnail

Step 2: Extract what you can without downloading

Always try to get free data first (no disk needed):

Installs
1
First Seen
1 day ago