Video DeltaNet: A Video-Native Hybrid Attention for Livestream Video Generation

Hugging Face Daily Papers Papers
video-generation hybrid-attention diffusion-models linear-attention efficiency

Summary

Video DeltaNet presents a hybrid attention mechanism combining Softmax and linear attention to enhance efficiency in video generation models, achieving a 14.5x speedup over baseline methods.

Video diffusion models repeatedly process long spatiotemporal token sequences during denoising, making attention a major computational bottleneck. Linear attention offers an appealing alternative and has been widely adopted in recent large language models, but directly applying it to video models often fails to preserve the fine-grained interactions required for high-quality generation. We present Video DeltaNet (VDN), which combines local Softmax attention with bidirectional linear memory for long-range video context. Its linear branch introduces Video Delta Attention (VDA), which updates memory once per frame by jointly incorporating its 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. We instantiate VDN 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, corresponding to a 14.5x speedup over the 50-step dense H3 baseline on the same GPU count.
Original Article
View Cached Full Text

Cached at: 09/18/26, 03:00 AM

Paper page - Video DeltaNet: A Video-Native Hybrid Attention for Livestream Video Generation

Source: https://huggingface.co/papers/2609.20744 Authors:

,

,

,

,

,

,

,

,

,

Abstract

Videodiffusionmodelsrepeatedlyprocesslongspatiotemporaltokensequencesduringdenoising,makingattentionamajorcomputationalbottleneck.Linearattentionoffersanappealingalternativeandhasbeenwidelyadoptedinrecentlargelanguagemodels,butdirectlyapplyingittovideomodelsoftenfailstopreservethefine-grainedinteractionsrequiredforhigh-qualitygeneration.WepresentVideoDeltaNet(VDN),whichcombineslocalSoftmaxattentionwithbidirectionallinearmemoryforlong-rangevideocontext.ItslinearbranchintroducesVideoDeltaAttention(VDA),whichupdatesmemoryonceperframebyjointlyincorporatingitsspatialtokens.Separateoutputprojectionsandlearnablegatescalibratethetwobranches,whileastagedteacher-alignmentrecipeprogressivelyintroducesthenewpathwayintopretrainedmodels.WeinstantiateVDNonMiniMaxH3,applyingthehybridtovideo-to-videointeractionswhileretainingSoftmaxforinteractionsinvolvingtextoraudio.Witheight-stepdistillationandanoptimizedSGLangservingstack,VDN-H3completesDiTdenoisingfora14.3-second,768pvideoin6.70secondsoneightNVIDIAB200GPUs,correspondingtoa14.5xspeedupoverthe50-stepdenseH3baselineonthesameGPUcount.

View arXiv pageView PDFProject pageAdd to collection

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2609.20744 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2609.20744 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2609.20744 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

OpenVDN/vdn-minimax-h3

Hugging Face Models Trending

VDN-Minimax-H3 is an open-source hybrid-attention model that speeds up video generation with near-lossless quality, featuring fast inference and plug-and-play adapters powered by MiniMax H3.

FVAttn: Adaptive Sparse Attention with Runtime Load Balancing for Video Generation

Hugging Face Daily Papers

FVAttn is a training-free sparse attention system that uses runtime load balancing to improve distributed execution efficiency of adaptive sparse attention under multi-GPU sequence parallelism for video generation, achieving up to 4.41x attention speedup and 2.11x inference speedup over FlashAttention on step-distilled Wan2.2 I2V while maintaining competitive video quality.

Nvidia's New Long-Form Video Generation (12 minute read)

TLDR AI

NVIDIA Research introduces SANA-Video 2.0, a hybrid video diffusion transformer that generates high-quality 720p video on a single GPU, achieving up to 120× speedup over Wan 2.2-14B via hybrid linear-softmax attention and block attention residuals.