WorldFlow3D: Flowing Through 3D Distributions for Unbounded World Generation

2026-03-30

AI Summary

This work addresses the fundamental challenge of generating unbounded 3D worlds by transcending the limitations of conventional conditional denoising approaches. It introduces a latent-space-free, streaming generation framework that formulates 3D synthesis as a continuous transport process between data distributions. Leveraging flow matching, the method directly produces geometrically accurate and spatially coherent 3D structures. By integrating vectorized scene layouts with visual texture control mechanisms, the framework enables efficient and controllable generation. Evaluated on both real-world outdoor driving sequences and synthetic indoor environments, the proposed approach substantially outperforms existing methods for unbounded scene generation, achieving significant improvements in generation fidelity, convergence speed, and cross-domain generalization capability.

Abstract

Unbounded 3D world generation is emerging as a foundational task for scene modeling in computer vision, graphics, and robotics. In this work, we present WorldFlow3D, a novel method capable of generating unbounded 3D worlds. Building upon a foundational property of flow matching - namely, defining a path of transport between two data distributions - we model 3D generation more generally as a problem of flowing through 3D data distributions, not limited to conditional denoising. We find that our latent-free flow approach generates causal and accurate 3D structure, and can use this as an intermediate distribution to guide the generation of more complex structure and high-quality texture - all while converging more rapidly than existing methods. We enable controllability over generated scenes with vectorized scene layout conditions for geometric structure control and visual texture control through scene attributes. We confirm the effectiveness of WorldFlow3D on both real outdoor driving scenes and synthetic indoor scenes, validating cross-domain generalizability and high-quality generation on real data distributions. We confirm favorable scene generation fidelity over approaches in all tested settings for unbounded scene generation. For more, see https://light.princeton.edu/worldflow3d.

Problem

• unbounded 3D world generation
• 3D scene modeling
• 3D data distributions
• scene generation fidelity
• cross-domain generalizability

Innovation

• flow matching
• unbounded 3D generation
• latent-free 3D modeling
• controllable scene synthesis
• 3D distribution transport