# motionmaxxing

> An agent skill that captures a brand from a website and builds, renders and checks motion-graphics films using measured motion timing and scripted quality gates.

motionmaxxing is an open-source agent skill by Tejas Makwana for making launch films, product promos, brand stings and kinetic-type pieces. It runs in Claude Code and in any agent that reads a SKILL.md folder, and renders locally from plain HTML using Chrome and ffmpeg. The README describes it as an early release.

## What It Is

The skill is a folder with a SKILL.md and a toolbox of scripts that guides an agent through a ten-step protocol. Given a website, it captures the brand, finds an idea for the film, builds a world and storyboard, and animates with timing measured from 53 professional motion moments. It then renders the film and inspects it with scripts rather than simply asserting that it came out well.

## How the Workflow Runs

The agent shows the idea and storyboard before building. Steps cover brand capture, finding the idea, world and time planning, optional script and voice, building the hardest beat first, an honest look at the render, sound, and delivery. When a gate fails, it is fixed in a fixed order.

Outputs per film:
- film/final.mp4, with sound when an ElevenLabs key is set
- film/index.html, editable HTML source using the Motion runtime
- film/STORYBOARD.md, the plan and beat table
- film/NOTE.md, an account of the idea, what is real versus generated, and the gate numbers

## Measured Motion and Quality Gates

The README says about 4,300 moves and 12 named eases were measured from the studied moments. Findings include a median move of 10 frames, landing that decelerates and leaving that accelerates, and arrival from oversize rather than from zero.

Quality gates G0 to G5 check that the render exists, there are no empty frames, the end card is short, and there is no page chrome. G0, G2, G3 are computed by look.py and G5 by lint.mjs. G1 and G4 are judged by eye on a contact sheet.

## Toolbox and Runtime

Scripts handle brand capture, official logo fetching, precedent lookup, generated surface plates, ElevenLabs voice, sound effects and music, word-to-frame sync, a loudness-targeted mix, deterministic frame-by-frame rendering, and blind A/B review. The runtime, motion.js, builds on GSAP with seek-safe frames, and includes a Three.js hero3d object and native iOS UI furniture.

## Setup and Known Limits

Requires Node 22+, Python 3, ffmpeg and Google Chrome; an ElevenLabs key and Codex image generation are optional. Documented limits include that the lint cannot see text inside a canvas or image, WebGL renders run at about 7-11 fps, and ElevenLabs music does not always hit tempo or length. The skill is Apache-2.0 licensed, with bundled third-party components under their own licenses.

## Features
- Brand capture from a website URL
- Ten-step film production protocol
- 12 measured named eases
- Deterministic frame-by-frame rendering through Chrome and ffmpeg
- Scripted quality gates G0-G5 including page-chrome lint
- Contact sheet, cut and end-hold analysis
- ElevenLabs voice-over, sound effects and music
- Word-timing sync and loudness-targeted audio mix
- Three.js hero3d objects and native iOS UI furniture
- Generated surface plates with a guard against text, logos and people
- Storyboard and honest NOTE.md output per film
- Blind A/B review page generator

## Integrations
Claude Code, ElevenLabs, GSAP, Three.js, ffmpeg, Google Chrome, Codex CLI, FinalMotion

## Platforms
IOS, API, CLI

## Pricing
Open Source

## Links
- Website: https://github.com/Tejashmakwana/motionmaxxing
- Documentation: https://github.com/Tejashmakwana/motionmaxxing/blob/main/runtime/README.md
- Repository: https://github.com/Tejashmakwana/motionmaxxing
- EveryDev.ai: https://www.everydev.ai/tools/motionmaxxing