academa-labs/manimgx

The animation engine for agents: blazingly fast math and 3D videos with Manim's Python API, on a Rust GPU core.

What it solves

ManimGX is a high-performance 3D animation engine designed to create mathematical and scientific animations. It solves the problem of slow render times associated with traditional animation tools like ManimCE, while maintaining compatibility with the Manim Community Edition (CE) API, making it a drop-in replacement for many users.

How it works

Built with Rust and powered by wgpu, the engine leverages native GPU rendering and WebGPU for browser-based execution. It uses Typst for typesetting text and math (translating LaTeX input via mitex), and x264 for video encoding. For browser execution, it utilizes Pyodide to run Python scenes in a Web Worker.

Who it’s for

  • AI Agents: The project is specifically optimized for agents (e.g., Claude Code, Cursor) to write Python code that renders into high-quality 3D animations quickly.
  • Educators and Mathematicians: Those using the Manim API to create visual explanations of complex concepts.
  • Web Developers: Users who want to embed 3D animations in web pages without a render server.

Highlights

  • Extreme Speed: Up to 90x faster than ManimCE in benchmarks.
  • API Compatibility: Compatible with Manim CE's API, allowing existing scenes to be adapted easily.
  • Browser Support: Can be run directly in the browser via npm install manimgx.
  • Integrated Tooling: Includes built-in support for LaTeX/Typst math, synchronized speech (self.say()), and an inspect mode for storyboarding.
  • Static Typing: Fully statically typed, including keyword arguments and .animate chains.

Related

  • Project
    WebODM/OpenSplat

    OpenSplat is a C++ implementation of 3‑D Gaussian splatting that converts COLMAP/OpenSfM/ODX/nerfstudio data into compact Gaussian‑based scene files. It supports CUDA, ROCm (HIP), Apple Metal, and CPU‑only builds, offers Docker images, and provides a simple CLI for training, resuming, masking, and exporting to .ply/.splat/.spz/.rad formats.

  • Project
    zju3dv/GVHMR

    GVHMR is a SIGGRAPH‑Asia 2024 research codebase that recovers world‑grounded 3‑D human motion from monocular video. It combines visual odometry (DPVO or SimpleVO) with a gravity‑aligned coordinate system, provides ready‑to‑run demo scripts, Colab/HuggingFace demos, and scripts to reproduce benchmark results on 3DPW, RICH, and EMDB.

  • Project
    NVlabs/nvdiffrec

    A differentiable rendering framework that extracts triangular 3D models, materials, and lighting from multi-view images through joint optimization.

  • Project
    Friedrich-M/UniMate

    UniMate is a diffusion‑based text‑to‑animation model that can generate motion for any rigged 3‑D skeleton (humans, animals, robots, etc.) without per‑skeleton fine‑tuning. It is trained on the UniML3D dataset (13.8 k clips, 7.4 k diverse skeletons) and provides full training, data‑processing, and inference pipelines, plus a tool to bring your own GLB/FBX/GLTF rigs into the system.