Edit yourvideos withClaude.

Set up this folder as my video editing workspace. Install these 4 free tools one at a time:
1. FFmpeg: https://ffmpeg.org
2. yt-dlp: https://github.com/yt-dlp/yt-dlp
3. Whisper: https://github.com/openai/whisper
4. HyperFrames: https://github.com/heygen-com/hyperframes (needs Node.js 22 or newer, install that too if it's missing)

Check my operating system first and use the right installer for it. After each tool, run its version command and show me the output. If one fails, fix it before you move on. Then create three folders: raw, refs and out. End with one line per tool: working or not.

These are reels from a creator whose editing I want to learn from:
[paste 3 to 5 reel links]

Download each one into refs with yt-dlp. For every reel, pull one frame every half second with FFmpeg and transcribe it with Whisper with word timestamps. Then write me a breakdown of the editing style: fonts (weight, size against the frame, colour, position), captions (words per line, how they animate, highlight colour), pacing (average shot length, where cuts land against the speech), zooms and motion, colour, sound design (music level, where sound effects hit), the first 3 seconds, and what this creator never does. Give numbers, not adjectives.

Turn that breakdown into one reusable file called STYLE.md in this folder. Write it as rules you will follow on every edit: the closest free font, sizes in px for a 1080x1920 frame, colours as hex codes, caption rules, cut rhythm, zoom rules, sound effect rules, safe margins and export settings (1080x1920, 30 fps, H.264). Add a short "never do" list. Keep it to one page. Use the style as a guide for my own footage, don't reuse their clips, logo or music. From now on, read STYLE.md before every edit.

New video: raw/[your file name]. I speak in [language, e.g. Hindi with English words]. Read STYLE.md first, then:
1. Transcribe it with Whisper with word timestamps.
2. Cut silences, filler words and retakes with FFmpeg. Keep the best take of each line.
3. Before you render anything, show me a storyboard: every shot with its timestamp, what is on screen, the caption text, any zoom or motion graphic, and any sound effect. Then wait for my notes.
4. When I approve, build the captions and motion graphics with HyperFrames, assemble with FFmpeg and export to out as a 1080x1920 MP4.
5. Send me 4 still frames from the export so I can check it before watching.