Wan2.1-Fun-1.3B-Control
| Property | Value |
|---|---|
| Model Size | 1.3B parameters |
| Storage Space | 19.0 GB |
| License | Apache License 2.0 |
| Author | alibaba-pai |
What is Wan2.1-Fun-1.3B-Control?
Wan2.1-Fun-1.3B-Control is a specialized video control model designed for advanced video manipulation and generation. It represents a significant advancement in controlled video synthesis, offering support for multiple control conditions and resolutions.
Implementation Details
The model is trained on 81 frames at 16 frames per second, supporting multiple resolutions including 512x512, 768x768, and 1024x1024. It incorporates various control mechanisms and supports multilingual prediction capabilities.
- Multi-resolution support (512, 768, 1024)
- Training configuration: 81 frames at 16fps
- Multiple control condition support
- Memory-efficient options including model_cpu_offload and qfloat8 quantization
Core Capabilities
- Control condition support (Canny, Depth, Pose, MLSD)
- Trajectory control functionality
- Multi-resolution video prediction
- Multilingual prediction support
- Memory optimization options for different GPU configurations
Frequently Asked Questions
Q: What makes this model unique?
The model's ability to handle multiple control conditions while supporting various resolutions and trajectory control makes it particularly versatile for video generation tasks. Its memory optimization options also make it accessible for different hardware configurations.
Q: What are the recommended use cases?
The model is ideal for controlled video generation tasks requiring specific visual elements like edge detection (Canny), depth information, pose estimation, or line detection (MLSD). It's particularly useful for applications requiring precise control over video generation parameters.
