https://github.com/haoz19/PhysRig

Physics-based rigging with MPM for realistic character animation. ICCV 2025.
https://github.com/haoz19/PhysRig

Last synced: 4 months ago
JSON representation

Physics-based rigging with MPM for realistic character animation. ICCV 2025.

• Host: GitHub
• URL: https://github.com/haoz19/PhysRig
• Owner: haoz19
• License: mit
• Created: 2024-09-27T01:38:20.000Z (about 1 year ago)
• Default Branch: main
• Last Pushed: 2025-07-09T20:42:14.000Z (4 months ago)
• Last Synced: 2025-07-10T06:51:18.833Z (4 months ago)
• Homepage: https://physrig.github.io/
• Size: 4.88 KB
• Stars: 30
• Watchers: 4
• Forks: 0
• Open Issues: 1
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

• ai-game-devtools - PhysRig - Based Rigging for Realistic Articulated Object Modeling. |[arXiv](https://arxiv.org/abs/2506.20936) | | Model | (<span id="model">3D Model</span> / <span id="tool">LLM (LLM & Tool)</span>)

README

          
# PhysRig: Differentiable Physics-Based Rigging for Realistic Articulated Object Modeling

[![Paper](https://img.shields.io/badge/Paper-arXiv%3A2506.20936-b31b1b.svg)](https://arxiv.org/abs/2506.20936)

[![Project Page](https://img.shields.io/badge/Project-physrig.github.io-blue)](https://physrig.github.io)

[![ICCV 2025](https://img.shields.io/badge/ICCV-2025-red)](https://iccv2025.thecvf.com/)

PhysRig is a differentiable, physics-based skinning and rigging framework that enables **realistic deformation** of articulated 3D objects. Unlike traditional methods like Linear Blend Skinning (LBS), PhysRig embeds skeletons into a deformable soft-body volume simulated using **Material Point Method (MPM)** — capturing the behavior of soft tissues, tails, ears, and other elastic structures in a physically plausible way.

> 🔧 From UIUC & Stability AI | 🦖 ICCV 2025 Accepted

---

## 🧠 Highlights

- **Differentiable Physics-Based Simulation**  

  Deformations are modeled using MPM and continuum mechanics, supporting gradient-based optimization of both motion and material properties.

- **Material Prototypes**  

  Learnable elastic parameter templates (Young’s modulus & Poisson’s ratio) provide expressive yet compact material modeling.

- **Driving Point Rigging**  

  Instead of bone transformations, PhysRig uses velocity-driven embedded points to induce dynamic shape changes.

- **Strong Performance**  

  Outperforms LBS on both user studies and geometric metrics (Chamfer Distance), with better convergence and physical realism.

- **Applications**  

  - Inverse Skinning (from motion to parameters)  

  - Pose Transfer (e.g., human → jellyfish)  

  - 4D Reconstruction & Animation

---

## Code will be released to the public around ICCV