Kinetic Editorial Skill — Development skill for Claude Code

kinetic-editorial-reel

A [Claude Code](https://claude.com/claude-code) skill that turns a topic, script or voiceover into a vertical (9:16) **kinetic editorial reel**. Every spoken word appears on screen as it's said, with one red keyword per line and a literal hero object for each phrase. Scenes alternate between warm cream paper and black glow, with slam hits, camera pushes and sound effects. It's built with [HyperFrames](https://hyperframes.heygen.com) and Three.js.

What the skill does

1. Intake: asks for the topic or message, the voiceover and transcript, the edit speed (calm / standard / rapid), the assets and the sound (SFX only, or with music).
2. Storyboard: writes a frame-by-frame plan, with an image-generation prompt list if assets are missing. You approve it before anything is built.
3. Preparation: subagents handle two jobs in parallel:
   • Generate an offline sound-effect pack (scripts/make_sfx.py).
   • Measure anchor points on your images so props land exactly where they should.
   • Meanwhile the main session prepares assets, word timings and fonts:
     • Assets: transparent PNGs are used as-is. Normal images get plain backgrounds removed (scripts/prep_assets.py).
     • Word timings: estimated from the voice's loudness envelope, with no speech-to-text model (scripts/word_timing.py).
4. Build: one HyperFrames composition with word-by-word captions, a GSAP timeline and Three.js layers where 3D helps. Sound effects are placed automatically (scripts/place_sfx.py).
5. Verify and preview: lint, snapshots, hyperframes check, then the Studio preview. It renders only when you say so.

Requirements

• Claude Code
• The HyperFrames skills: npx hyperframes skills update general-video
• Node.js (for npx hyperframes), ffmpeg, and Python 3 with numpy, scipy, Pillow: pip install numpy scipy pillow

Install

Clone into a skills folder:

# for one project only
git