Playable Environments: Video Manipulation in Space and Time

We present Playable Environments-a new representation for interactive video generation and manipulation in space and time. With a single image at inference time, our novel framework allows the user to move objects in 3D while generating a video by providing a sequence of desired actions. The actions are learnt in an unsupervised manner. The camera can be controlled to get the desired viewpoint. Our method builds an environment state for each frame, which can be manipulated by our proposed action module and decoded back to the image space with volumetric rendering. To support diverse appearances of objects, we extend neural radiance fields with style-based modulation. Our method trains on a collection of various monocular videos requiring only the estimated camera parameters and 2D object locations. To set a challenging benchmark, we introduce two large scale video datasets with significant camera movements. As evidenced by our experiments, playable environments enable several creative applications not attainable by prior video synthesis works, including playable 3D video generation, stylization and manipulation(1).

Automated summary

We present Playable Environments, a novel representation for interactive video generation and manipulation. It allows the user to move objects in 3D and generate videos by providing a sequence of desired actions based on a single image at inference time. The actions are learned in an unsupervised manner and the camera can be controlled to achieve the desired viewpoint. Our method builds an environment state for each frame that can be manipulated using our proposed action module and rendered back to the image space with volumetric rendering. We also introduce two large-scale video datasets with significant camera movements to set a challenging benchmark. Playable environments enable creative applications such as 3D video generation, stylization, and manipulation that were not possible with prior video synthesis works.