Wan-Move Examples
Explore the capabilities of motion-controllable video generation
Motion Control Capabilities
Wan-Move generates high-quality 5-second videos at 832×480p resolution with precise point-level motion control. Each example demonstrates how dense point trajectories guide the movement of objects within the generated videos. The framework supports a wide range of motion control applications, from simple single-object movements to complex multi-object choreography.
The examples below showcase different capabilities of the framework. Each application demonstrates how latent trajectory guidance enables precise control over motion while maintaining high video quality and natural appearance.
Single-Object Motion Control
Precise control over individual object movements
Control the movement of individual objects within a scene. Define trajectories that specify exactly where objects should move, and Wan-Move generates videos where those objects follow the specified paths while maintaining natural appearance and interaction with the environment. This application is useful for creating animated scenes where a single element needs to move in a specific way.
Key Characteristics
- ▸Precise control over individual object movements with point-level accuracy
- ▸Natural motion that respects physics and scene context
- ▸Maintains video quality throughout the video duration
- ▸Proper interaction with lighting, shadows, and environment
Multi-Object Motion Control
Choreograph multiple objects simultaneously
Choreograph multiple objects simultaneously, each following independent trajectories. This capability enables complex scenes where different elements move in coordinated or independent patterns, creating dynamic compositions. Each object can have its own trajectory, allowing for sophisticated animations with multiple moving parts that interact naturally within the scene.
Key Characteristics
- ▸Independent trajectory control for multiple objects in the same scene
- ▸Coordinated or independent motion patterns for different elements
- ▸Complex scene dynamics with multiple moving elements
- ▸Natural interaction between moving objects
Camera Control
Simulate professional camera movements
Simulate professional camera movements without physically moving a camera. Wan-Move supports various camera operations including linear displacement, dolly in, dolly out, panning, and other cinematic movements. This creates professional-looking camera work that would traditionally require specialized equipment and expertise.
Key Characteristics
- ▸Linear displacement for smooth tracking shots and follow movements
- ▸Dolly in and out for depth-based camera movement and zoom effects
- ▸Panning for horizontal scene exploration and reveal shots
- ▸Professional cinematic camera movements without physical equipment
Motion Transfer
Reuse motion patterns across different content
Extract motion patterns from existing videos and apply them to different content. This allows for reusing successful motion templates across different scenes and subjects, enabling consistent motion styles. Motion transfer is particularly useful for creating series of videos with similar motion characteristics or applying proven motion patterns to new content.
Key Characteristics
- ▸Extract motion patterns from reference videos using trajectory tracking
- ▸Apply extracted motion to new content with different subjects
- ▸Reuse successful motion templates for consistent styling
- ▸Efficient content creation through motion pattern reuse
3D Rotation
Generate videos with 3D object rotation
Generate videos showing objects rotating in three-dimensional space. This is particularly useful for product demonstrations, architectural visualization, and any application requiring 360-degree views of objects or scenes. The 3D rotation capability enables viewers to see objects from multiple angles in a single video.
Key Characteristics
- ▸Full 360-degree rotations in three-dimensional space
- ▸Ideal for product demonstrations and showcase presentations
- ▸Architectural and object visualization from multiple angles
- ▸Smooth rotation with proper perspective and lighting changes
How Wan-Move Works
Latent Trajectory Guidance Process
The core technique in Wan-Move is latent trajectory guidance. This method represents motion conditions by propagating features from the first frame along user-defined trajectories.
- 1.Provide Initial Frame: Start with a single image showing the starting state of the scene with all objects in their initial positions
- 2.Define Point Trajectories: Specify where elements should move by providing x,y coordinates for each point across frames
- 3.Feature Propagation: The model propagates first-frame features along these trajectories through the latent space
- 4.Video Generation: Generate video content that respects trajectory constraints while maintaining quality and natural motion
Input Requirements
- ▸Initial Image: Starting frame with scene appearance
- ▸Trajectory Data: NumPy arrays with point coordinates
- ▸Visibility Masks: Occlusion information for points
- ▸Text Prompt: Scene description for context
Output Specifications
- ▸Resolution: 832×480 pixels
- ▸Duration: 5 seconds
- ▸Output Fidelity: High-quality realistic appearance
- ▸Motion Accuracy: Precise trajectory following
Evaluation on MoveBench
MoveBench provides standardized test cases for evaluating motion-controllable video generation. The benchmark includes both single-object and multi-object scenarios across diverse content categories. Each test case includes reference images, trajectory annotations, visibility masks, and text descriptions in both English and Chinese.
Researchers can use MoveBench to objectively compare different approaches and measure performance. The benchmark enables fair evaluation of motion accuracy, temporal consistency, and video quality across different methods.
Performance Characteristics
Qualitative comparisons show that Wan-Move produces videos with motion accuracy comparable to commercial solutions. The framework has been compared with both academic methods and commercial solutions, demonstrating competitive performance in motion controllability while being open for research and development.
Video Quality Metrics
Try Wan-Move
To experiment with Wan-Move, you can install it locally following the instructions on the installation page. The system requires Python with PyTorch 2.4.0 or later, downloaded model weights, and GPU hardware for practical inference times.
Gradio Demo Coming Soon
The research team has indicated plans to release a Gradio demo interface that will provide a user-friendly way to interact with Wan-Move. This demo will allow users to upload images, define trajectories through an interactive interface, and generate videos without writing code.