Send a Loom video to Claude, GPT, or Gemini

Steps

1. Get the Loom .mp4 URL

   Open the Loom share link, click the … menu → Download Original. Or use yt-dlp.

   yt-dlp -f best -o 'loom.%(ext)s' 'https://www.loom.com/share/...'

2. Install peepshow

   npm install -g peepshow

3. Run extraction

   peepshow ./loom.mp4 --emit json > run.json

4. Hand to your LLM

   Drag the frame folder into Claude / ChatGPT, or pipe JSON to an API call.

   peepshow ./loom.mp4 --sink slack   # or --sink notion / obsidian

Why it works

Loom is the default async-video tool. Recordings are usually 1-15 minutes — long enough that Gemini native video upload starts adding up but short enough that scene-detection captures most signal in ~15 frames. peepshow + whisper.cpp turns a Loom drop into a structured artifact teammates can grep.

When it helps

• Async standups and design reviews captured in Loom.
• Bug repros where a screen-recording explains what a paragraph can't.
• Sales / customer-success libraries — every Loom auto-indexed for retrieval.

Pitfalls

• Loom private links require auth — Loom's official API or workspace cookies for yt-dlp.
• Audio quality varies — whisper.cpp's `base` model handles most narration, `large-v3` for noisy environments.

Works with these LLMs

Pairs with these sinks