Edit a video with your voice using Claude and HyperFrames (prompts included)

You are my video editor. You use HyperFrames to build the edit as code.
Here is my talking-head video: [file path or name].
1. Transcribe it word by word with the timing of each word.
2. Find every command I give out loud (right hand, left hand, reframe me, center me, split screen, show me…).
3. Cut the bad takes and long silences, and show me the list of cuts before applying them.
Don't create any effect yet.

When I say "show me the world", darken the image behind me and show a world map with every country.
When I say "break it", make the map explode: each country flies in a different direction, with a flash and an impact sound synced on the word.
When I say "rebuild Africa", bring only the African countries back into place, color them gold and zoom in.
The rest of the world stays in pieces, very transparent.

When I say "in my right hand", make a card appear floating just above my palm, with a golden glow under the card.
The card follows my hand if I move it. Measure the position of my palm frame by frame.
Careful: the video is mirrored, my right hand is on the right of the screen.
Right hand card content: [your number or visual].
Left hand card content: [your number or visual].
Make the number count up.

When I say "reframe me to the left", slide the image so I'm on the left of the screen, and bring the animation in on my right, at the same height as me. No split screen.
When I say "reframe me to the right", do the opposite.
When I say "center me", go back to the center.
Animation on the left: [e.g. a 3D house that builds floor by floor, then blows apart into layers with a label on each layer].
Animation on the right: [e.g. a building whose floors get reorganized by AI].

When I say "split screen", split the screen in two: me at the bottom, and at the top [e.g. a rotating 3D globe with my cities: Lagos, Abidjan, Niamey, Dakar, Nairobi, Paris, Montreal, Dubai].
When I name two cities, connect them with a glowing arc.

Add transition sound effects, with these rules:
1. Each sound must land exactly on the word that triggers the effect. Sync the PEAK of the sound to the word, not the start of the file.
2. Loud sounds (impacts, whooshes) only when I'm not talking.
3. While I'm talking: only small, short, soft sounds.
4. Automatically lower the sounds whenever my voice is present.
Show me the list of sounds with their timing before rendering.

Make two versions: a vertical 9:16 one for Reels and TikTok, and a horizontal 16:9 one for YouTube and LinkedIn.
In vertical, rethink each scene so it stays readable: nothing gets cut off, the visuals stay on my hands, and the reframes stay side by side.
Add word-by-word captions, with the key words in gold.