ExCellGen: Fast, Controllable, Photorealistic 3D Scene Generation from a Single Real-World Exemplar

TL;DR

ExCellGen is a rapid, controllable framework that generates photorealistic 3D scenes from a single real-world exemplar using 3D reconstruction and generative modeling, enabling interactive editing within minutes.

Contribution

We introduce a novel, fast pipeline combining 3D Gaussian Splatting and a generative cellular automaton for controllable 3D scene synthesis from minimal input.

Findings

Scene reconstruction in under 10 minutes

Scene generation in 0.5-2 seconds

Interactive editing with user control

Abstract

Photorealistic 3D scene generation is challenging due to the scarcity of large-scale, high-quality real-world 3D datasets and complex workflows requiring specialized expertise for manual modeling. These constraints often result in slow iteration cycles, where each modification demands substantial effort, ultimately stifling creativity. We propose a fast, exemplar-driven framework for generating 3D scenes from a single casual input, such as handheld video or drone footage. Our method first leverages 3D Gaussian Splatting (3DGS) to robustly reconstruct input scenes with a high-quality 3D appearance model. We then train a per-scene Generative Cellular Automaton (GCA) to produce a sparse volume of featurized voxels, effectively amortizing scene generation while enabling controllability. A subsequent patch-based remapping step composites the complete scene from the exemplar's initial 3D Gaussian splats, successfully recovering the appearance statistics of the input scene. The entire pipeline can be trained in less than 10 minutes for each exemplar and generates scenes in 0.5-2 seconds. Our method enables interactive creation with full user control, and we showcase complex 3D generation results from real-world exemplars within a self-contained interactive GUI.