Turbo4DGen: Ultra-Fast Acceleration for 4D Generation

4D generation, or dynamic 3D content generation, integrates spatial, temporal, and view dimensions to model realistic dynamic scenes, playing a foundational role in advancing world models and physical AI. However, maintaining long-chain consistency across both frames and viewpoints through the unique spatio-camera-motion (SCM) attention mechanism introduces substantial computational and memory overhead, often leading to out-of-memory (OOM) failures and prohibitive generation times. To address these challenges, we propose Turbo4DGen, an ultra-fast acceleration framework for diffusion-based multi-view 4D content generation. Turbo4DGen introduces a spatiotemporal cache mechanism that persistently reuses intermediate attention across denoising steps, combined with dynamically semantic-aware attention pruning and an adaptive SCM chain bypass scheduler, to drastically reduce redundant SCM attention computation. Our experimental results show that Turbo4DGen achieves an average 9.7$\times$ speedup without quality degradation on the ObjaverseDy and Consistent4D datasets. To the best of our knowledge, Turbo4DGen is the first dedicated acceleration framework for 4D generation.

Generating realistic 3D content that changes over time is important for applications such as virtual environments, robotics, and AI systems that model the physical world. However, current methods are often computationally expensive because they must maintain consistency across both time and camera viewpoints. We introduce Turbo4DGen, an acceleration framework for dynamic 3D content generation. By reusing intermediate information and selectively reducing unnecessary computation, Turbo4DGen substantially improves efficiency while preserving generation quality. Experiments show an average 9.7× speedup without quality degradation, making high-quality dynamic 3D generation more practical.