Search papers, labs, and topics across Lattice.
Resonance4D introduces a novel framework for physics-driven 4D dynamic simulation that combines 3D Gaussian Splatting with the Material Point Method, supervised by a new Dual-domain Motion Supervision (DMS) technique. DMS enforces dynamic consistency by combining spatial structural consistency with frequency-domain spectral consistency, avoiding the need for computationally expensive online video diffusion or optical-flow pipelines. The framework also incorporates zero-shot text-prompted segmentation and simulation-guided initialization to enable stable full-parameter physical recovery, achieving high fidelity and motion consistency with reduced memory overhead.
Forget expensive video diffusion: Resonance4D unlocks high-fidelity 4D dynamic simulations by cleverly supervising motion in the frequency domain.
Physics-driven 4D dynamic simulation from static 3D scenes remains constrained by an overlooked contradiction: reliable motion supervision often relies on online video diffusion or optical-flow pipelines whose computational cost exceeds that of the simulator itself. Existing methods further simplify inverse physical modeling by optimizing only partial material parameters, limiting realism in scenes with complex materials and dynamics. We present Resonance4D, a physics-driven 4D dynamic simulation framework that couples 3D Gaussian Splatting with the Material Point Method through lightweight yet physically expressive supervision. Our key insight is that dynamic consistency can be enforced without dense temporal generation by jointly constraining motion in complementary domains. To this end, we introduce Dual-domain Motion Supervision (DMS), which combines spatial structural consistency for local deformation with frequency-domain spectral consistency for oscillatory and global dynamic patterns, substantially reducing training cost and memory overhead while preserving physically meaningful motion cues. To enable stable full-parameter physical recovery, we further combine zero-shot text-prompted segmentation with simulation-guided initialization to automatically decompose Gaussians into object-part-level regions and support joint optimization of full material parameters. Experiments on both synthetic and real scenes show that Resonance4D achieves strong physical fidelity and motion consistency while reducing peak GPU memory from over 35\,GB to around 20\,GB, enabling high-fidelity physics-driven 4D simulation on a single consumer-grade GPU.
