PhysForge: Generating Physics-Grounded 3D Assets for Interactive Virtual World

TL;DR

PhysForge is a novel framework that generates physics-grounded 3D assets for virtual worlds by combining hierarchical physical planning with a diffusion model, supported by a large annotated dataset.

Contribution

It introduces PhysForge, a two-stage generative approach utilizing a large-scale dataset and a new KineVoxel Injection mechanism for realistic, functional 3D asset synthesis.

Findings

PhysForge produces functionally plausible 3D assets ready for simulation.

The framework effectively integrates hierarchical physical constraints into asset generation.

Experiments validate the realism and utility of generated assets for virtual environments.

Abstract

Synthesizing physics-grounded 3D assets is a critical bottleneck for interactive virtual worlds and embodied AI. Existing methods predominantly focus on static geometry, overlooking the functional properties essential for interaction. We propose that interactive asset generation must be rooted in functional logic and hierarchical physics. To bridge this gap, we introduce PhysForge, a decoupled two-stage framework supported by PhysDB, a large-scale dataset of 150,000 assets with four-tier physical annotations. First, a VLM acts as a "physical architect" to plan a "Hierarchical Physical Blueprint" defining material, functional, and kinematic constraints. Second, a physics-grounded diffusion model realizes this blueprint by synthesizing high-fidelity geometry alongside precise kinematic parameters via a novel KineVoxel Injection (KVI) mechanism. Experiments demonstrate that PhysForge produces functionally plausible, simulation-ready assets, providing a robust data engine for interactive 3D content and embodied agents.