{"slug": "dynamic-in-few-step-unifying-dynamic-computation-and-few-step-distillation-for", "title": "Dynamic-in-Few-Step: Unifying Dynamic Computation and Few-Step Distillation for Efficient Video Generation", "summary": "Researchers propose Dynamic-in-Few-Step, a post-training acceleration framework that integrates dynamic structural sparsification into few-step distillation for video diffusion models. The method transforms a pre-trained VDM into a step-specific Mixture-of-Models, achieving a 30x speedup over the 50-step teacher on Wan-14B while preserving generation quality. This approach addresses varying computational demands across denoising stages and is orthogonal to existing acceleration techniques.", "body_md": "arXiv:2607.06631v1 Announce Type: new\nAbstract: Video Diffusion Models (VDMs) have demonstrated superior generation quality but suffer from prohibitive computational costs. While recent few-step distillation techniques significantly accelerate inference, they typically enforce a static model architecture across all denoising stages, ignoring the varying computational demands inherent to different noise levels. In this work, we propose a novel post-training acceleration framework that exploits this redundancy by integrating dynamic structural sparsification directly into the distillation process. Unlike conventional post-hoc compression applied to a fixed diffusion pipeline, our approach jointly optimizes the denoising steps and structured model sparsity, transforming a pre-trained VDM into a compact, step-specific Mixture-of-Models (MoM). To address the training instability arising from this joint optimization, we introduce a Progressive Training Strategy coupled with an Output Rollout Mechanism, which ensures the coherent learning of structural decisions across timesteps. Furthermore, we develop a specialized inference engine to deploy the resulting MoM efficiently. Our method is orthogonal to existing acceleration techniques and highly effective: On Wan-14B, it removes 24% of the per-step FLOPs on top of 4-step distillation, adding a 1.2x wall-clock gain and reaching a 30x speedup over the 50-step teacher while preserving competitive generation quality.", "url": "https://wpnews.pro/news/dynamic-in-few-step-unifying-dynamic-computation-and-few-step-distillation-for", "canonical_source": "https://arxiv.org/abs/2607.06631", "published_at": "2026-07-09 04:00:00+00:00", "updated_at": "2026-07-09 04:28:31.536015+00:00", "lang": "en", "topics": ["machine-learning", "generative-ai", "ai-infrastructure", "computer-vision", "artificial-intelligence"], "entities": ["Wan-14B"], "alternates": {"html": "https://wpnews.pro/news/dynamic-in-few-step-unifying-dynamic-computation-and-few-step-distillation-for", "markdown": "https://wpnews.pro/news/dynamic-in-few-step-unifying-dynamic-computation-and-few-step-distillation-for.md", "text": "https://wpnews.pro/news/dynamic-in-few-step-unifying-dynamic-computation-and-few-step-distillation-for.txt", "jsonld": "https://wpnews.pro/news/dynamic-in-few-step-unifying-dynamic-computation-and-few-step-distillation-for.jsonld"}}