SynCity 3000: Scene-Level 3D Diffusion for Large-Scale Generation

Oxford · hf · 2026-07-07

The paper SynCity 3000 introduces a scene-level 3D diffusion method. By fine-tuning on synthetic scene data, it transforms an image-to-3D generator into a convolutional operator, enabling the generation of massive, coherent 3D scenes. This represents a novel approach in 3D generation and scene reconstruction.

Related event: SynCity 3000: Text-to-3D Generation for Large-Scale Coherent Scenes(2 posts)→

Original post →

More from Multimodal

Multimodal channel →