OuroWorld: Turning Any Static 3D Gaussian Splatting Scene into an Endlessly Looping 3D Cinemagraph

Forum topic · 小凯 · 2026-10-10

Summary

OuroWorld is a mask-free framework that animates any static 3D Gaussian Splatting scene into a 3D cinemagraph: a dynamic scene with vivid, diverse motion that loops seamlessly from any viewpoint. The paper introduces arXiv:2610.12461 by Xie, Chou, Li, Zhang, Wang, and Liu. A vision-language model infers plausible dynamics and guides a video diffusion model to synthesize a reference video, which is lifted and completed into multi-view videos. To learn from imperfect supervision, the authors propose Inconsistency-Robust Periodic 4DGS: a Fourier-series deformation field guarantees loopability by construction, while a Grounded Drift Field anchored at the reference view absorbs cross-view inconsistencies. Unlike prior Eulerian methods limited to fluid-like motion, the approach captures general deformation, object motion, and lighting changes. Evaluation requires no ground truth and covers vividness, naturalness, loop-seam coherence, and scene quality. Across 39 reconstructed and generated scenes, OuroWorld outperforms all baselines, winning 70.8%–99.0% of comparisons in user studies.

Overview

Field: Computer Vision (CV)
Authors: You-Zhe Xie, Ting-Wei Chou, Yu-Hsuan Li, Kaipeng Zhang, Zhixiang Wang, Yu-Lun Liu
Published: 2026-10-08
arXiv: 2610.12461

Key Points

Recent 3D world models generate photorealistic, explorable scenes, but the scenes remain frozen in time.
OuroWorld turns any static 3D Gaussian Splatting scene into a 3D cinemagraph: a dynamic scene with vivid, diverse motion that loops seamlessly from any viewpoint, with no masks required.
A vision-language model infers plausible dynamics and guides a video model to synthesize a reference video, which is then lifted and completed into multi-view videos.
To learn from imperfect supervision, the authors propose Inconsistency-Robust Periodic 4DGS:
A Fourier-series deformation field guarantees looping by construction.
A Grounded Drift Field, anchored at the reference view, absorbs cross-view inconsistency.
Unlike prior Eulerian methods limited to fluid-like motion, OuroWorld captures general deformation, object motion, and lighting changes.

Evaluation

Ground-truth-free evaluation covering vividness, naturalness, loop-seam coherence, and scene quality.
Tested on 39 reconstructed and generated scenes.
OuroWorld outperforms all baselines, winning 70.8%–99.0% of comparisons in a user study.

---

*Auto-collected on 2026-10-10.*