CubeComposer: Spatio-Temporal Autoregressive 4K 360° Video Generation from Perspective Video

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Li, Lingen, Wang, Guangzhi, Li, Xiaoyu, Zhang, Zhaoyang, Dou, Qi, Gu, Jinwei, Xue, Tianfan, Shan, Ying
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866915834286309376
author Li, Lingen
Wang, Guangzhi
Li, Xiaoyu
Zhang, Zhaoyang
Dou, Qi
Gu, Jinwei
Xue, Tianfan
Shan, Ying
author_facet Li, Lingen
Wang, Guangzhi
Li, Xiaoyu
Zhang, Zhaoyang
Dou, Qi
Gu, Jinwei
Xue, Tianfan
Shan, Ying
contents Generating high-quality 360° panoramic videos from perspective input is one of the crucial applications for virtual reality (VR), whereby high-resolution videos are especially important for immersive experience. Existing methods are constrained by computational limitations of vanilla diffusion models, only supporting $\leq$ 1K resolution native generation and relying on suboptimal post super-resolution to increase resolution. We introduce CubeComposer, a novel spatio-temporal autoregressive diffusion model that natively generates 4K-resolution 360° videos. By decomposing videos into cubemap representations with six faces, CubeComposer autoregressively synthesizes content in a well-planned spatio-temporal order, reducing memory demands while enabling high-resolution output. Specifically, to address challenges in multi-dimensional autoregression, we propose: (1) a spatio-temporal autoregressive strategy that orchestrates 360° video generation across cube faces and time windows for coherent synthesis; (2) a cube face context management mechanism, equipped with a sparse context attention design to improve efficiency; and (3) continuity-aware techniques, including cube-aware positional encoding, padding, and blending to eliminate boundary seams. Extensive experiments on benchmark datasets demonstrate that CubeComposer outperforms state-of-the-art methods in native resolution and visual quality, supporting practical VR application scenarios. Project page: https://lg-li.github.io/project/cubecomposer
format Preprint
id arxiv_https___arxiv_org_abs_2603_04291
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle CubeComposer: Spatio-Temporal Autoregressive 4K 360° Video Generation from Perspective Video
Li, Lingen
Wang, Guangzhi
Li, Xiaoyu
Zhang, Zhaoyang
Dou, Qi
Gu, Jinwei
Xue, Tianfan
Shan, Ying
Computer Vision and Pattern Recognition
Artificial Intelligence
Generating high-quality 360° panoramic videos from perspective input is one of the crucial applications for virtual reality (VR), whereby high-resolution videos are especially important for immersive experience. Existing methods are constrained by computational limitations of vanilla diffusion models, only supporting $\leq$ 1K resolution native generation and relying on suboptimal post super-resolution to increase resolution. We introduce CubeComposer, a novel spatio-temporal autoregressive diffusion model that natively generates 4K-resolution 360° videos. By decomposing videos into cubemap representations with six faces, CubeComposer autoregressively synthesizes content in a well-planned spatio-temporal order, reducing memory demands while enabling high-resolution output. Specifically, to address challenges in multi-dimensional autoregression, we propose: (1) a spatio-temporal autoregressive strategy that orchestrates 360° video generation across cube faces and time windows for coherent synthesis; (2) a cube face context management mechanism, equipped with a sparse context attention design to improve efficiency; and (3) continuity-aware techniques, including cube-aware positional encoding, padding, and blending to eliminate boundary seams. Extensive experiments on benchmark datasets demonstrate that CubeComposer outperforms state-of-the-art methods in native resolution and visual quality, supporting practical VR application scenarios. Project page: https://lg-li.github.io/project/cubecomposer
title CubeComposer: Spatio-Temporal Autoregressive 4K 360° Video Generation from Perspective Video
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2603.04291