OmniRoam: World Wandering via Long-Horizon Panoramic Video Generation

TL;DR

OmniRoam is a panoramic video generation framework enabling long-horizon scene wandering with high fidelity, controllability, and consistency, leveraging a novel two-stage process and new datasets.

Contribution

The paper introduces a controllable panoramic video generation framework with a preview and refine stage, and provides new datasets for training and evaluation.

Findings

Outperforms state-of-the-art in visual quality and consistency

Enables high-resolution, long-range scene wandering

Supports real-time generation and 3D reconstruction extensions

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.