Full-4D: Generating Full-Scope 4D Scenes from a Single-View Video

A research team has proposed a framework to generate full-scope 4D scenes from a single-view video, addressing a problem long considered ill-posed due to the lack of multi-angle information in a single recording. The work, submitted on 25 May 2026, tackles the challenge of reconstructing a complete, dynamic 3D scene over time from a single camera feed [1]. The authors note that existing methods are typically limited to monocular videos, simple 3D effects, or only small viewpoint perturbations around the original viewpoint, falling short of true 4D generation [2]. A core obstacle has been the absence of large-scale datasets that capture full-scope 4D scenes with synchronized multi-view videos [2]. To address this, the team introduces Real-MV-4D, a large-scale dataset of synchronized multi-view videos captured in diverse real-world environments to provide the necessary 4D supervision [2]. The framework casts full-scope 4D generation as a two-stage process: multi-view video synthesis followed by optimization-based 4D reconstruction from the generated views [2]. A multi-view video diffusion model is trained with a novel fused time-view attention mechanism that directly embeds geometric reprojection priors and explicit camera conditioning into its view-time interactions [2]. Unlike basic feature fusion, this direct binding strictly aligns the generation process with physical 3D priors to produce a dense, synchronized time-by-view video grid [2]. The synthesized multi-view videos are then lifted into an explicit 4D representation, specifically 4D Gaussian Splatting, and regularized by a Flow Matching Distillation loss that exploits the multi-view prior to improve novel-view rendering [2]. The concept of synthesizing multiple viewpoints from limited input data echoes historical efforts in stereoscopic imaging. Active shutter 3D systems, for instance, present alternating left-eye and right-eye images so rapidly that the brain fuses them into a single 3D image, using liquid crystal shutter glasses that become opaque when voltage is applied [4]. The broader history of film technology shows that 3D film techniques have existed from the beginning of cinema but only became a standard option in most movie theaters during the first decades of the 21st century [5]. The researchers report that extensive experiments demonstrate their method outperforms existing approaches in both visual fidelity and geometric consistency, enabling full-scope 4D scene generation from single-view videos [2]. The paper is hosted on arXiv, and associated code and data are linked through platforms including Hugging Face and ScienceCast [1].