Bridging the Embodiment Gap: Disentangled Cross-Embodiment Video Editing
For robot learning researchers, this work offers a scalable method to leverage abundant human videos for training robots, addressing a key data bottleneck.
The paper addresses the distribution shift between human and robot videos for robotic manipulation learning by proposing a generative framework for cross-embodiment video editing that learns disentangled task and embodiment representations. The method generates coherent robot execution videos from single human demonstrations without paired data, achieving temporally consistent and morphologically accurate results.
Learning robotic manipulation from human videos is a promising solution to the data bottleneck in robotics, but the distribution shift between humans and robots remains a critical challenge. Existing approaches often produce entangled representations, where task-relevant information is coupled with human-specific kinematics, limiting their adaptability. We propose a generative framework for cross-embodiment video editing that directly addresses this by learning explicitly disentangled task and embodiment representations. Our method factorizes a demonstration video into two orthogonal latent spaces by enforcing a dual contrastive objective: it minimizes mutual information between the spaces to ensure independence while maximizing intra-space consistency to create stable representations. A parameter-efficient adapter injects these latent codes into a frozen video diffusion model, enabling the synthesis of a coherent robot execution video from a single human demonstration, without requiring paired cross-embodiment data. Experiments show our approach generates temporally consistent and morphologically accurate robot demonstrations, offering a scalable solution to leverage internet-scale human video for robot learning.