Video Creation by Demonstration

TL;DR

This paper introduces a novel method called $ ext{delta}$-Diffusion for generating realistic continuation videos from a demonstration video and a context image, enabling flexible and expressive video creation.

Contribution

It presents $ ext{delta}$-Diffusion, a self-supervised training approach that learns from unlabeled videos and uses implicit latent control for flexible video generation.

Findings

$ ext{delta}$-Diffusion outperforms baselines in human preference and machine evaluations.

The method enables interactive world simulation.

It effectively extracts action latents with minimal appearance leakage.

Abstract

We explore a novel video creation experience, namely Video Creation by Demonstration. Given a demonstration video and a context image from a different scene, we generate a physically plausible video that continues naturally from the context image and carries out the action concepts from the demonstration. To enable this capability, we present $\delta$-Diffusion, a self-supervised training approach that learns from unlabeled videos by conditional future frame prediction. Unlike most existing video generation controls that are based on explicit signals, we adopts the form of implicit latent control for maximal flexibility and expressiveness required by general videos. By leveraging a video foundation model with an appearance bottleneck design on top, we extract action latents from demonstration videos for conditioning the generation process with minimal appearance leakage. Empirically, $\delta$-Diffusion outperforms related baselines in terms of both human preference and large-scale machine evaluations, and demonstrates potentials towards interactive world simulation. Sampled video generation results are available at https://delta-diffusion.github.io/.