Find your next open-source project

Search by name, purpose, or keyword — or paste a GitHub repository URL

artemnovitckii/claude-animation-skills

Media & Design

Claude Code skills that render short vertical videos—3D explainers, stick figures, line art, routes, exploded views, and graphs—locally from text prompts.

115
15
JavaScript
NOASSERTION
1 day old

What it does

This repository packages six Claude Code skills that turn a short text prompt into a vertical animation for Reels, without requiring footage. Each skill has a fixed visual style: a 3D cartoon explainer (/zach-d-films), a stick-figure explainer (/stick-figure), minimal black-and-white line art (/minimalistic-graphics), a 3D map route (/map-route), a 3D exploded view of an object (/breakdown), and a minimal line graph (/graph) L1–17.

The skills are split into two groups. The four "quick" skills work from a small JSON spec: Claude writes the spec, runs the skill's build.py, and produces an MP4 plus a contact sheet with Instagram's safe area marked. The two "engine" skills are full explainer reels where Claude copies an example project, writes the script with the user, handles voiceover and caption timing, and renders the result L45–61. Animation is drawn with GSAP for the 2D skills and three.js for the 3D ones, and rendered through HyperFrames and headless Chrome via Playwright L62–76.

Quick start

You need macOS or Linux with ffmpeg, Node.js 18+, and Python 3.9+ L18–26:

brew install ffmpeg node
git clone https://github.com/artemnovitckii/claude-animation-skills.git ~/claude-animation-skills
cd ~/claude-animation-skills && ./install.sh

install.sh links the skills into ~/.claude/skills/, installs three.js and Playwright into kit/node_modules, fetches GSAP, and warms up HyperFrames. For /stick-figure and /zach-d-films, also run ./install.sh --voice to download about 1 GB of local voice and timing models; no API key is required L32–34.

Then open Claude Code and use a skill directly:

/graph 0 to 100K followers in 9 months: 1.2K, 2.1K, 3.4K, 5K, 9K, 24K, 51K, 78K, 100K
/map-route home to the gym, 2.8 km, 9 minutes

For a quick skill you get name.<skill>.mp4 at 1080x1920, 30 fps, plus a contact sheet; the README estimates about 10 seconds for 2D and about 1.5 minutes for 3D. The engine skills are a working session rather than a single command L49–56. Live examples are on the project page.

Alternatively, install it as a Claude Code plugin with /plugin marketplace add artemnovitckii/claude-animation-skills and /plugin install animation-skills@claude-animation-skills; plugin skills are namespaced (e.g. /animation-skills:graph), and the README notes the git clone route is the one tested end to end L39–42.

Constraints

  • macOS or Linux only; ffmpeg, Node.js 18+, and Python 3.9+ are required L18–26.
  • The quick skills and engine skills have different workflows: spec-based rendering versus a longer scripted explainer session L45–61.
  • Voiceover and word timings run locally and need about 1 GB of models downloaded once; they are not bundled in the repository L32–34.
  • Dependencies, including HyperFrames, three.js, GSAP, and Playwright, are installed at setup rather than included. The README labels HyperFrames as being developed by HeyGen L77–85.
  • /zach-d-films is named after the style of short it produces and is not affiliated with Zack D. Films L77–85.
  • The repository is MIT licensed, but the GitHub license metadata is listed as NOASSERTION.

Extending

To add a quick skill, copy a folder such as skills/graph, change template.html (reading the spec from const SPEC = __SPEC__), update its DEFAULTS in build.py, and rerun ./install.sh L64–73.

HotRepo's take

This repository was created on 2026-11-01 and first appears in HotRepo's new-project snapshot on 2026-11-02 for the JavaScript list: 111 stars and 15 forks, with 7 new stars recorded in a roughly two-hour observation window. The daily all-language rank is not known, so the data supports only that burst of early attention on the new list, not a broader trend. The more durable point is structural: instead of shipping a model, the project packages reusable prompting and rendering skills around existing open-source animation and local voice tooling, which keeps output generation on the user's machine.

Written by AI from the project's README, updated . The project's own docs are the source of truth.