Groundbreaking AI Technique Enables Novel Camera Trajectories for In-the-Wild Videos

ReCamMaster, a groundbreaking AI technique, enables generating novel camera trajectories for in-the-wild videos through video conditioning, backed by the MultiCamVideo dataset and open-source code, pre-trained models, and training instructions for the Wan2.1 model.

• ReCamMaster is a method for re-capturing in-the-wild videos with novel camera trajectories through a proposed video conditioning scheme.
• A multi-camera synchronized video dataset called MultiCamVideo, rendered with Unreal Engine 5, is released.
• The open-source code, pre-trained models, and instructions for inference and training on the Wan2.1 model are provided.