AsymRnR: Video Diffusion Transformers Acceleration with Asymmetric Reduction and Restoration

HunyuanVideo +AsymRnR (30% faster) (original) avg. latency -703s avg. latency -913s prompt: A person is knitting ... FastVideo-Hunyuan +AsymRnR (24% faster) (original) avg. latency -128s avg. latency -159s prompt: A teddy bear is swimming in the ocean ... Figure 1. Quality and speed comparison between baseline models, HunyuanVideo (Team, 2024c) and FastVideo-Hunyuan (Team, 2024a), with our AsymRnR. Our approach enables training-free, lossless acceleration for state-of-the-art video diffusion transformers.