Wan-Move: Motion-controllable Video Generation via Latent Trajectory Guidance
Ruihang Chu, Yefei He, Zhekai Chen, Shiwei Zhang, Xiaogang Xu, Bin Xia, Dingdong Wang, Hongwei Yi, Xihui Liu, Hengshuang Zhao, Yu Liu, Yingya Zhang, Yujiu Yang
2025-12-10
Summary
This paper introduces Wan-Move, a new method for controlling the motion in videos created by artificial intelligence. It's a way to tell the AI exactly how you want things to move within the video.
What's the problem?
Currently, controlling motion in AI-generated videos is difficult. Existing methods are often clunky, offering only basic control and not scaling well to create complex or high-quality movements. It's hard to get the AI to move things *precisely* how you want them to, and making improvements often requires significant changes to the underlying AI model.
What's the solution?
Wan-Move solves this by focusing on making the AI's understanding of the initial video frame 'aware' of motion. It does this by tracking the movement of objects as a series of points, then translating those movements into a special 'feature map' that guides the video creation process. This map essentially tells each part of the scene how to move. Importantly, Wan-Move doesn't require changing the core AI model used to generate the video; it simply adds this motion guidance on top, making it easier to implement and improve.
Why it matters?
This work is important because it allows for much more precise and realistic motion control in AI-generated videos. The results are comparable to professional tools, and the method is scalable, meaning it can be used to create longer, higher-resolution videos. The researchers also created a new dataset, MoveBench, to help others evaluate and improve motion control techniques, furthering the field of AI video generation.
Abstract
We present Wan-Move, a simple and scalable framework that brings motion control to video generative models. Existing motion-controllable methods typically suffer from coarse control granularity and limited scalability, leaving their outputs insufficient for practical use. We narrow this gap by achieving precise and high-quality motion control. Our core idea is to directly make the original condition features motion-aware for guiding video synthesis. To this end, we first represent object motions with dense point trajectories, allowing fine-grained control over the scene. We then project these trajectories into latent space and propagate the first frame's features along each trajectory, producing an aligned spatiotemporal feature map that tells how each scene element should move. This feature map serves as the updated latent condition, which is naturally integrated into the off-the-shelf image-to-video model, e.g., Wan-I2V-14B, as motion guidance without any architecture change. It removes the need for auxiliary motion encoders and makes fine-tuning base models easily scalable. Through scaled training, Wan-Move generates 5-second, 480p videos whose motion controllability rivals Kling 1.5 Pro's commercial Motion Brush, as indicated by user studies. To support comprehensive evaluation, we further design MoveBench, a rigorously curated benchmark featuring diverse content categories and hybrid-verified annotations. It is distinguished by larger data volume, longer video durations, and high-quality motion annotations. Extensive experiments on MoveBench and the public dataset consistently show Wan-Move's superior motion quality. Code, models, and benchmark data are made publicly available.