OmniRoam: World Wandering via Long-Horizon Panoramic Video Generation

2026-03-31

AI Summary

Existing video generation models predominantly rely on perspective views, which often compromise scene completeness and long-term spatiotemporal consistency, thereby limiting immersive navigation experiences. To address this, this work proposes OmniRoam, a novel framework that introduces panoramic video representation into controllable long-duration generation, enabling high-fidelity scene roaming through trajectory-based control. The approach operates in two stages: first generating a panoramic overview, followed by joint temporal extension and spatial super-resolution. Trained on a newly curated dataset comprising both synthetic and real-world panoramic videos, OmniRoam significantly outperforms existing methods in visual quality, controllability, and long-term consistency. Moreover, it supports practical extensions such as real-time generation and 3D reconstruction.

Abstract

Modeling scenes using video generation models has garnered growing research interest in recent years. However, most existing approaches rely on perspective video models that synthesize only limited observations of a scene, leading to issues of completeness and global consistency. We propose OmniRoam, a controllable panoramic video generation framework that exploits the rich per-frame scene coverage and inherent long-term spatial and temporal consistency of panoramic representation, enabling long-horizon scene wandering. Our framework begins with a preview stage, where a trajectory-controlled video generation model creates a quick overview of the scene from a given input image or video. Then, in the refine stage, this video is temporally extended and spatially upsampled to produce long-range, high-resolution videos, thus enabling high-fidelity world wandering. To train our model, we introduce two panoramic video datasets that incorporate both synthetic and real-world captured videos. Experiments show that our framework consistently outperforms state-of-the-art methods in terms of visual quality, controllability, and long-term scene consistency, both qualitatively and quantitatively. We further showcase several extensions of this framework, including real-time video generation and 3D reconstruction. Code is available at https://github.com/yuhengliu02/OmniRoam.

Problem

• panoramic video generation
• scene modeling
• long-horizon consistency
• global consistency
• video synthesis

Innovation

• panoramic video generation
• long-horizon scene wandering
• trajectory-controlled generation
• spatiotemporal consistency
• two-stage refinement