Video diffusion models repeatedly process long spatiotemporal token sequences during denoising, which makes attention a major computational bottleneck. Linear attention has become a popular alternative in recent large language models, but the paper notes that applying it directly to video models often fails to preserve the fine-grained interactions needed for high-quality generation.
To address this, the authors present Video DeltaNet (VDN), a video-native hybrid attention approach that combines local Softmax attention with bidirectional linear memory for long-range video context. Its linear branch, called Video Delta Attention (VDA), updates memory once per frame by jointly incorporating the frame's spatial tokens.
Separate output projections and learnable gates calibrate the two branches, while a staged teacher-alignment recipe progressively introduces the new pathway into pretrained models. VDN is instantiated on MiniMax H3, applying the hybrid to video-to-video interactions while retaining Softmax for interactions involving text or audio.
With eight-step distillation and an optimized SGLang serving stack, VDN-H3 completes DiT denoising for a 14.3-second, 768p video in 6.70 seconds on eight NVIDIA B200 GPUs. According to the paper, this corresponds to a 14.5x speedup over the 50-step dense H3 baseline on the same GPU count.