Event ·Multi-source

视频生成模型 Video DeltaNet 提升直播视频生成效率

视频生成模型 Video DeltaNet 通过结合局部 Softmax 注意力和双向线性内存，解决了视频扩散模型在去噪过程中处理长时空序列时的计算瓶颈问题。其线性分支引入的 Video Delta Attention (VDA) 每帧更新一次记忆，同时整合空间令牌。在 MiniMax H3 模型上应用后，VDN-H3 在八块 NVIDIA B200 GPU 上完成 14.3 秒 768p 视频的去噪仅需 6.70 秒，比 50 步密集 H3 基准快 14.5 倍。

Current read

这个新方法能显著提升视频生成速度，对需要快速生成直播视频的场景很有用。

8 sources44° AI heatLast updated 2026/9/17 17:33:15

Evidence

8 sources

Primary sourcearXiv cs.LG

2026/9/17 17:33:15View source

Related source 1NVIDIA AI

2026/9/16 01:04:21View source

Related source 2Gary Marcus

2026/9/16 04:22:08View source

Related source 3Tech in Asia

2026/9/16 08:02:53View source

Related source 4IT之家

2026/9/17 05:04:39View source

Related source 5techcrunch

2026/9/17 14:06:14View source

Related source 6SEA-LION (AI Singapore)

2026/9/18 06:52:08View source

Related source 7elvis

2026/9/18 22:24:02View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.

Open the story