ReCamMaster: A Framework for Generating Novel Views and Motion Trajectories from a Single Video 133

What is ReCamMaster?

ReCamMaster is a framework developed jointly by research teams from Zhejiang University, Kuaishou Technology, the Chinese University of Hong Kong, and Huazhong University of Science and Technology for generating new viewpoints and camera-trajectory videos from a single video. It can generate video content with new viewpoints and motion trajectories from a single video by redefining the camera trajectory. The core of this technology lies in leveraging a pretrained text-to-video generation model, combined with a carefully designed video conditioning mechanism, to achieve high-quality video re-rendering.

ReCamMaster Features

• Camera trajectory control: Supports a variety of complex trajectories such as panning, rotation, and zoom, allowing users to customize camera movements.

• High-quality video generation: Maintains dynamic synchronization and multi-frame consistency across the video, producing high-quality output.

• Large-scale dataset support: A multi-camera synchronized video dataset built using Unreal Engine 5, covering diverse scenes and camera motions.

• Innovative video conditioning mechanism: Significantly improves the quality of generated videos through frame-dimensional condition concatenation.

ReCamMaster Use Cases

• Video stabilization: Converts shaky footage into smooth video while preserving the original scene and motion.

• Video super-resolution and outpainting: Enhances video resolution and expands the field of view.

• Autonomous driving and robotics vision: Provides multi-view data augmentation to improve model generalization.

• Video creation and post-production: Redefines the camera trajectory of a video to create new visual effects.

How to Use ReCamMaster

1. Input video: Provide a single-view input video.

2. Select a trajectory: Choose or customize a camera trajectory (such as pan, rotate, zoom, etc.).

3. Generate video: The system generates a new-viewpoint video based on the defined trajectory and delivers the result via email.

4. Batch processing: For batch processing or comparative testing, contact the project team (jianhongbai@zju.edu.cn).

GitHub repository:https://github.com/KwaiVGI/ReCamMaster

Project page:https://jianhongbai.github.io/ReCamMaster/

Paper:https://arxiv.org/abs/2503.11647