This custom Claude skill literally watches video frame by frame.

This custom Claude skill watches video frame by frame. Not just the transcript, the actual visuals. Because in any good video, half the meaning is happening on screen, not in the words. Now Claude sees both. Here's how you can use it. Say a YouTube video goes viral. You just type slash Claude watch, paste the URL, hit enter.

YouTube video goes viral. You just type slash Claude watch, paste the URL, hit enter. Claude studies the whole thing frame by frame and writes you a full video breakdown. A 30 minute video broken down and rewritten in under 2 minutes. You can also do this with every top creator in your niche. Paste the URL, Claude breaks down

do this with every top creator in your niche. Paste the URL, Claude breaks down what's working, the hooks, the visuals, the trends, and pipes it straight into your obsidian vector library. Your second brain just keeps getting smarter on autopilot. You are never out of content ideas. For the full setup tutorial, check my YouTube channel in my

out of content ideas. For the full setup tutorial, check my YouTube channel in my profile. Now here's how it works under the hood. Claude doesn't have a native video model, so the skill splits every video into the two things. Claude does understand, images and text. It uses YTDLP to pull the video from

any site on the internet. YouTube, Loom, Instagram and over all platforms. Then FFmpeg rips out frames across the whole video and pulls a clean audio track. The audio gets transcribed, free if YouTube has captions or through Whisper on Grok's free tier if not. Then everything gets handed to Claude with the timestamps lined up. So

Claude isn't guessing from a transcript anymore. It's flipping through the frames a flip book while reading the script. It knows exactly what's on screen the moment something is being said. All running locally. No expensive video API. AI to get this custom created Claude skill.

AI to get this custom created Claude skill.