⚙️ The Reverse Motion System. Pixel-perfect motion design from any reference

FREEA Koda guide

Free guide: pixel-perfect motion design from any reference. Reverse the motion you love into a spec your agent rebuilds shot for shot. By @timkoda_

Drag across · scroll

stop prompting blind measure the reference spec becomes prompt render move for three ways to

GETTHESYSTEM

Everything you need, in one place

The kit

The Reverse Motion System

The measuring skill + spec pipeline for your agent. Works with Claude Code, Cursor, Codex, and 40+ agents.

npx skills add timkoda/reverse-motion-systemGet the skill

The kit

Higgsfield MCP

The render engine. Connects Claude Code to Seedance, Fable 5 and 30+ models.

claude mcp add --transport http --scope user higgsfield https://mcp.higgsfield.ai/mcpWire the MCP

Built with

The Reverse Motion System

Computer vision measures the motion you love, frame by frame: cuts, camera, shapes, palette. The spec becomes a structured prompt your agent renders move for move. Copy it pixel-perfect, the

Get the skill

Time: 40 minutes
Level: Comfortable in a terminal
Tools: Higgsfield · Claude Code · Seedance
You end with: A motion piece rendered move for move

Every camera move, shape and colour is a measured value. Recreate any motion design pixel perfect, then flip the system and render your own.

🔗

Resources

get the system

Free preview

The rest of this guide is free.

Create a free account and keep reading right here. Takes 30 seconds, no card.

Create a free account

Already have one? Log in

One skill, one connector. Install both and the whole pipeline is live.

Requires Claude Code and Python 3.10+ with OpenCV (pip install opencv-python-headless numpy).

The Shift

stop prompting blind

Everyone prompts video models blind: type an adjective, pray, regenerate. The result never matches the film in your head, because the model is improvising every value you left undefined. The Reverse Motion System removes the guessing. Computer vision MEASURES a reference video frame by frame: every cut, every camera move, every shape, every colour. Those numbers become a spec file, the spec becomes a structured JSON prompt, and the Higgsfield MCP renders it move for move. Copy a reference to prove the control. Then feed it your own shapes, fonts and camera moves, and render a piece that is entirely yours.