AI Video Editor: every prompt from the reel (Claude Code)

AI Video Editor: every prompt from the reel

Edit videos with Claude Code, no course needed. Run these in order.

Before you start

1. Download the Claude desktop app and open the Code tab (not Claude in the browser).
2. On your computer, make a folder called AI Video Editor with three folders inside: raw footage, references and exports.
3. Start a new Claude Code chat, choose Local, and select the AI Video Editor folder.

Prompt 1: install the five tools

Set up this folder as my video editing workspace. Install and test these five tools:

1. Remotion, to build the video and the motion graphics in code
2. HyperFrames, for HTML-based motion graphics
3. FFmpeg, to cut, mix and export video
4. Whisper, to transcribe my footage with word-level timestamps for captions
5. yt-dlp, to download reference videos from a link

Install anything missing, then run a small test for each one (render a 3-second test clip,
transcribe 5 seconds of audio, download one short public video) and tell me when all five work.
Write what you installed and how to run each tool to a file called SETUP.md in this folder.

Prompt 2: teach it your taste

Paste 3 to 5 links to videos in the style you want.

Here are reference videos in the style I want for my edits:

[link 1]
[link 2]
[link 3]

Download each one into the references folder with yt-dlp. Study them and write a STYLE.md file
covering: pacing (average shot length, how often something new appears), captions (font, size,
position, how words appear), colour and contrast, motion graphics and transitions, music and sound
effects, and how each video opens in the first 3 seconds.

Make my edits inspired by them, not a copy. Keep my face, my voice and my brand.

Prompt 3: edit the video

Film like you normally do. If you mess up a line, say "hey editor, cut that" and keep going. Drop the file into raw footage, then run:

Edit the newest file in raw footage, following STYLE.md.

1. Transcribe it with Whisper (word-level timestamps).
2. Cut the dead air and long pauses.
3. Whenever I say "hey editor, cut that", remove that take and the line I said just before it.
4. Add captions in my style and music that fits, ducked under my voice.
5. Export a vertical 9:16 version into exports.

Then show me a few stills and a list of every cut you made, so I can review it before you
render the final version.

Something off? Say it in plain words, for example: "make the hook bigger", "the captions are too small", "cut the second pause after the intro".

Prompt 4 (optional): no camera, use your clone

Clone your voice on ElevenLabs and make a HeyGen avatar of yourself first, then:

Write a 60-second script for a reel about [topic], in the way I talk (use my past transcripts in
this folder as the example). Voice it with my ElevenLabs voice clone and render my HeyGen avatar
saying it. Then edit it following STYLE.md, the same way as my filmed videos.

Prompt 5: save it as a skill

This edit is right. Save everything you did, the steps, the settings and STYLE.md, as a skill
called my-reel-edit, so next time I can drop a file in raw footage and run one prompt.

Prompt 6: post it

Connect Pozy to Claude (pozy.ai), connect your Instagram, TikTok and YouTube, then:

Post the newest video in exports to my Instagram, TikTok and YouTube with a caption in my style.
On Instagram, turn on comment to DM so everyone who comments gets [your link] in their DMs.