WonderZoom: Multi-Scale 3D World Generation

TL;DR

WonderZoom introduces a multi-scale 3D scene generation method from a single image, using scale-adaptive surfels and progressive detail synthesis to produce coherent scenes at varying granularities, outperforming existing models.

Contribution

It proposes a novel scale-aware 3D representation and synthesis approach enabling multi-scale scene generation from a single image, addressing limitations of prior single-scale models.

Findings

Outperforms state-of-the-art models in quality and alignment

Enables interactive zooming and detail synthesis in 3D scenes

Successfully generates scenes with details from landscapes to microscopic features

Abstract

We present WonderZoom, a novel approach to generating 3D scenes with contents across multiple spatial scales from a single image. Existing 3D world generation models remain limited to single-scale synthesis and cannot produce coherent scene contents at varying granularities. The fundamental challenge is the lack of a scale-aware 3D representation capable of generating and rendering content with largely different spatial sizes. WonderZoom addresses this through two key innovations: (1) scale-adaptive Gaussian surfels for generating and real-time rendering of multi-scale 3D scenes, and (2) a progressive detail synthesizer that iteratively generates finer-scale 3D contents. Our approach enables users to "zoom into" a 3D region and auto-regressively synthesize previously non-existent fine details from landscapes to microscopic features. Experiments demonstrate that WonderZoom significantly outperforms state-of-the-art video and 3D models in both quality and alignment, enabling multi-scale 3D world creation from a single image. We show video results and an interactive viewer of generated multi-scale 3D worlds in https://wonderzoom.github.io/