SS4D: Native 4D Generative Model via Structured Spacetime Latents

TL;DR

SS4D is a novel 4D generative model that synthesizes dynamic 3D objects from monocular videos, emphasizing high fidelity, temporal coherence, and structural consistency through structured spacetime latents.

Contribution

It introduces a direct 4D training approach with structured spacetime latents, leveraging pre-trained 3D models and temporal compression for efficient, coherent dynamic scene generation.

Findings

Achieves high-quality, temporally coherent 4D synthesis from monocular videos.

Utilizes a pre-trained 3D model to enhance spatial consistency.

Employs temporal compression for efficient long-sequence inference.

Abstract

We present SS4D, a native 4D generative model that synthesizes dynamic 3D objects directly from monocular video. Unlike prior approaches that construct 4D representations by optimizing over 3D or video generative models, we train a generator directly on 4D data, achieving high fidelity, temporal coherence, and structural consistency. At the core of our method is a compressed set of structured spacetime latents. Specifically, (1) To address the scarcity of 4D training data, we build on a pre-trained single-image-to-3D model, preserving strong spatial consistency. (2) Temporal consistency is enforced by introducing dedicated temporal layers that reason across frames. (3) To support efficient training and inference over long video sequences, we compress the latent sequence along the temporal axis using factorized 4D convolutions and temporal downsampling blocks. In addition, we employ a carefully designed training strategy to enhance robustness against occlusion