FracGen generates realistic crack and tear videos from single images
FracGen: Learning How Objects Stretch and Tear with Physics-Informed Video Generation
Computer Vision and Pattern Recognition
Summary
Predicting how objects break or tear when stressed is hard, especially from just one picture. The authors created FracGen, a tool that learns to generate believable videos of objects cracking and stretching by training on simulated fracture data. This simulation, called FracSim, provides detailed maps about how damage spreads inside materials, helping FracGen learn the physics behind breaking. As a result, FracGen can produce videos showing realistic fractures that can be controlled by specifying how and where the object breaks, without needing expensive calculations each time.
What this means in practice
- •For visual effects artists: Create controllable, realistic breaking scenes for movies and games from a single image without manual simulation tuning.$Commercial implications: Enables studios to produce fracture animations faster and with less manual effort, improving CGI production pipelines.
- •For robotics simulation teams: Generate plausible visual damage sequences for testing robotic manipulation and material handling without running costly physics simulations.
Authors
Trong-Tung Nguyen, Jiahan Zhang, Anand Bhattad
Abstract
We introduce FracGen, a fracture-aware video generation model that produces plausible, controllable fracture dynamics from a single image of an intact object, conditioned on physics signals. To train FracGen, we build FracSim, a fracture-aware simulation framework that augments material point method (MPM) simulation with a continuum damage model, producing paired fracture videos and dense, pixel-aligned physical fields at no additional cost beyond standard rendering. FracGen leverages these maps in two ways: it is trained to jointly predict them alongside RGB video, encouraging the model to capture physical state rather than surface appearance; and it is supervised with physics-informed losses that encourage consistency among the predicted maps. As a result, FracGen captures distinct material-specific fracture behavior without expensive test-time simulation or per-scene tuning, while offering fine-grained control over where an object tears, how fast the crack propagates, and how much deformation precedes failure. We further introduce a benchmark for evaluating the physical plausibility of generated fracture video, and show through extensive experiments that FracGen outperforms existing video generation baselines in both physical and visual fidelity. Results are best viewed in our project website: https://fracgen.github.io/.