Search papers, labs, and topics across Lattice.
Reshoot-Anything introduces a self-supervised framework for video reshooting that overcomes the scarcity of multi-view training data by generating pseudo multi-view triplets from monocular videos. The method extracts smooth random-walk crop trajectories from a single video to serve as source and target views, with a synthetically generated anchor created by forward-warping the first frame of the source. This forces the model to learn 4D spatiotemporal structures for reconstructing the target view from the source, enabling state-of-the-art temporal consistency and novel view synthesis.
Training a video reshooting model on internet-scale monocular videos is now possible, thanks to a clever self-supervision trick that generates multi-view training data from a single video.
Precise camera control for reshooting dynamic videos is bottlenecked by the severe scarcity of paired multi-view data for non-rigid scenes. We overcome this limitation with a highly scalable self-supervised framework capable of leveraging internet-scale monocular videos. Our core contribution is the generation of pseudo multi-view training triplets, consisting of a source video, a geometric anchor, and a target video. We achieve this by extracting distinct smooth random-walk crop trajectories from a single input video to serve as the source and target views. The anchor is synthetically generated by forward-warping the first frame of the source with a dense tracking field, which effectively simulates the distorted point-cloud inputs expected at inference. Because our independent cropping strategy introduces spatial misalignment and artificial occlusions, the model cannot simply copy information from the current source frame. Instead, it is forced to implicitly learn 4D spatiotemporal structures by actively routing and re-projecting missing high-fidelity textures across distinct times and viewpoints from the source video to reconstruct the target. At inference, our minimally adapted diffusion transformer utilizes a 4D point-cloud derived anchor to achieve state-of-the-art temporal consistency, robust camera control, and high-fidelity novel view synthesis on complex dynamic scenes.
