Latent Space Jam: Layout Guided Video Synthesis

Computer Science > Computer Vision and Pattern Recognition arXiv:2610.30390 (cs) [Submitted on 21 Oct 2026] Title: Latent Space Jam: Layout Guided Video Synthesis Authors: Asaf Joseph, Moshe Mandel, Ohad Fried.

计算机科学 > 计算机视觉与模式识别 arXiv:2610.30390 (cs) [提交于 2026 年 10 月 21 日] 标题：Latent Space Jam: Layout Guided Video Synthesis（潜在空间大乱斗：布局引导的视频合成） 作者：Asaf Joseph, Moshe Mandel, Ohad Fried。

Abstract: Adding control to video generation often requires training, which demands additional computational resources, data, or architectural changes. We propose a training-free approach that utilizes pretrained video diffusion models to control object layout and motion using coarse layouts.

摘要：为视频生成增加控制通常需要训练，这需要额外的计算资源、数据或架构更改。我们提出了一种无需训练的方法，利用预训练的视频扩散模型，通过粗略布局来控制对象的布局和运动。

Our method operates in two passes: first, it steers spatiotemporal placement through prompt-guided cross-attention to produce a coarse visual guide; second, this guide conditions the same model to generate a high-quality, layout-consistent video.

我们的方法分两步运行：首先，通过提示引导的交叉注意力机制引导时空位置，生成粗略的视觉指南；其次，该指南作为条件输入同一模型，以生成高质量且布局一致的视频。

The approach enables coordinated motion with strong temporal consistency, supporting diverse trajectories and object interactions without additional training. Extensive experiments show that our method consistently outperforms prior work in overlap and trajectory accuracy, as well as visual quality across static, dynamic, and multi-object scene settings.

该方法实现了具有强时间一致性的协调运动，支持多种轨迹和对象交互，且无需额外训练。大量实验表明，我们的方法在重叠和轨迹准确性，以及静态、动态和多对象场景设置下的视觉质量方面，始终优于先前的工作。