hyperkidxp/hadi-got-motion

A motion graphics studio for Claude Code. You give it a concept (and, if you like, a title, a length and a script). It checks and installs its own tools, shows you 12 motion styles that each move their own way and 12 map looks as clickable thumbnails, tells you how long the film will take, then directs, builds and renders it: talking cut-out characters, kinetic type, maps, stat hits and title sequences. Or pick Let Claude decide and it takes the whole film end to end without another question. The rule is that none of it looks AI-made.
## ▶ Watch the example: Mr Pemberton's Mistake (34 s, sound on)
https://github.com/user-attachments/assets/ee35efc7-9e92-4437-ae2b-3dee4466b2b8
The grandfather paradox as a Victorian cut-out farce. It was built entirely with this skill from public-domain engravings, ElevenLabs voices and a 1914 ragtime record.
Full quality (1080×1920): mr-pembertons-mistake.mp4
Source: templates/examples/pemberton/
Beat by beat: references/case-pemberton.md
## Why it looks different
Real material only. Characters, props and backdrops come from public-domain and CC0 archives (or from ChatGPT images you choose to generate). Code cuts, hinges, times and lights them. It never draws the art.
A director's workflow. There's a brief, three concepts with the safest one killed, a style with five written rules, the voice as the clock, a contact sheet reviewed like a designer would, and one render.
Taste written down. Hook in the first 1.5 s, one hero per beat, paper that pops instead of fading, puppets on twos, real typefaces, a sound on every action, and silence before the punchline.
## Install
bash
git clone https://github.com/hyperkidxp/hadi-got-motion ~/.claude/skills/hadi-got-motion
Or download hadi-got-motion.zip from Releases and unzip it into ~/.claude/skills/.
Then open Claude Code and ask for something like "make a 30-second motion graphic about the grandfather paradox".
### What you need
Claude Code (desktop app or CLI) on macOS or Linux
Node.js 20+ (brew install node or nodejs.org) and Python 3.9+
An ElevenLabs account for voices (a free tier exists)
Optional: an OpenAI API key for ChatGPT image generation. Without one, Claude gives you ready-to-paste ChatGPT prompts.
The moment the skill starts, Claude checks your machine and lists anything missing at the top of the brief form; one click and it installs it with hgm install: FFmpeg, the canvas renderer, anime.js, three.js, the ElevenLabs SDK, Python imaging and the fonts. They go into one workspace (~/MotionStudio by default), and the hgm command goes on your PATH.
## How a session goes
1. Tools check. Before anything else, Claude checks your machine. The brief form opens with either a green "all installed" line or the list of what's missing and an install button.
2. The brief, by sight. One form: title, length, concept, then thumbnails of every motion style and map look to click, and your script, uploaded as a file or typed in (or have Claude write it, or go without a voiceover). Only the concept is required.
3. Time estimate. As soon as you submit, Claude tells you roughly how long the build will take and the few minutes it needs from you.