How to Use the Manim Video Agent Skill (manim-video)

What the manim-video skill makes the agent do

An agent skill is a folder of instructions and helper files that an AI agent loads when a task matches it. This skill covers mathematical and technical animations. Once it is installed, the agent follows a fixed pipeline: it writes a plan, generates a Python script that uses Manim to draw each scene, renders each scene to video, and joins the clips together with ffmpeg. Narration is optional.

The skill also includes design guidance on color palettes, animation timing, and typography so the scenes look consistent. Its creative standard asks for a narrative arc before any code, geometry shown before algebra, and a pause after each key reveal.

• Concept explainers, such as how neural networks learn
• Equation derivations, such as a step-by-step proof
• Algorithm visualizations, such as quicksort step by step
• Data stories with animated charts and counters
• Architecture diagrams that build up component by component

How do you install the manim-video skill?

Download the ZIP file of the skill from this page. Then install it in one of two ways. For a personal setup, unzip it into your ~/.claude/skills/ folder. For a single project, unzip it into that project's .claude/skills/ folder. You can also upload the ZIP in the Claude apps under Settings, then Capabilities, then Skills.

After installing, confirm the skill folder contains SKILL.md at its top level alongside the README and the references folder.

What do you need installed before rendering?

The skill needs Python, Manim Community Edition (installed with pip install manim), LaTeX, and ffmpeg. LaTeX is used to render equations through MathTex. On Linux the source names the texlive-full package, and on macOS it names mactex. LaTeX is a large download, so plan for that time and disk space.

The skill includes a setup script that checks these dependencies. The README shows how to run it from the skill folder. Check the README for the exact minimum Python version.

• Python (minimum version listed in the README)
• Manim Community Edition: pip install manim
• LaTeX: texlive-full on Linux, mactex on macOS
• ffmpeg
• Optional: ElevenLabs or Qwen3-TTS for narration

How do you trigger the skill with a natural request?

You do not need special commands. Describe the video you want in ordinary language. The skill is meant to trigger on requests for animated explanations, math animations, concept visualizations, algorithm walkthroughs, technical explainers, 3Blue1Brown-style videos, or programmatic animation with geometric or mathematical content.

Include the audience, the core idea, and the key moment you want the viewer to reach. The more specific the request, the better the plan.

• Explain how neural networks learn, for a first-year student, in about two minutes
• Animate the proof of the Pythagorean theorem
• Show how quicksort works step by step on an array of eight numbers
• Animate our before and after performance metrics as a short data story
• Show our microservice architecture building up, one service at a time

How does the agent build and render the video?

The pipeline runs in this order: plan, code, render, stitch, optional audio, and review. The agent writes plan.md with the narrative arc, scene list, and color palette. It writes script.py with one class per scene, and each scene can be rendered on its own. Draft renders use low quality for speed, and production renders use high quality.

Each scene is rendered to a clip and then joined into final.mp4 with ffmpeg. Manim writes its output under a media folder inside the project.

• plan.md: narrative arc, scene breakdown
• script.py: all scenes in one file
• concat.txt: the ffmpeg scene list
• final.mp4: the stitched output
• media/: created automatically by Manim

Safety notes and tips for better results

This skill runs scripts. The agent generates Python code and runs Manim and ffmpeg on your machine, so read the generated script before you run it on a project you care about. The listing states that the skill does not use the network. Optional narration uses ElevenLabs or Qwen3-TTS, which are separate tools with their own setup and terms, so check those before you enable narration.

For better results, ask for a single concept per video, state the audience, and review the draft render before asking for a production render. The skill's guidance favors clear, uncluttered frames over dense ones.

• Read generated scripts before running them on important projects
• Start with draft-quality renders to check timing and layout
• Name the audience and the one idea the viewer should leave with
• Keep text at a readable size; the skill sets a minimum font size of 18