PanoWan: Lifting Diffusion Video Generation Models to 360{\deg} with Latitude/Longitude-aware Mechanisms

TL;DR

PanoWan is a novel method that adapts pre-trained text-to-video models for high-quality 360-degree panoramic video generation by addressing spatial distortions and boundary issues, supported by a new panoramic video dataset.

Contribution

It introduces PanoWan, a lightweight framework with latitude-aware sampling and boundary handling mechanisms, enabling effective transfer of pre-trained models to panoramic video synthesis.

Findings

Achieves state-of-the-art panoramic video generation quality

Demonstrates robustness in zero-shot downstream tasks

Provides a new high-quality panoramic video dataset PanoVid

Abstract

Panoramic video generation enables immersive 360{\deg} content creation, valuable in applications that demand scene-consistent world exploration. However, existing panoramic video generation models struggle to leverage pre-trained generative priors from conventional text-to-video models for high-quality and diverse panoramic videos generation, due to limited dataset scale and the gap in spatial feature representations. In this paper, we introduce PanoWan to effectively lift pre-trained text-to-video models to the panoramic domain, equipped with minimal modules. PanoWan employs latitude-aware sampling to avoid latitudinal distortion, while its rotated semantic denoising and padded pixel-wise decoding ensure seamless transitions at longitude boundaries. To provide sufficient panoramic videos for learning these lifted representations, we contribute PanoVid, a high-quality panoramic video dataset with captions and diverse scenarios. Consequently, PanoWan achieves state-of-the-art performance in panoramic video generation and demonstrates robustness for zero-shot downstream tasks. Our project page is available at https://panowan.variantconst.com.