Moonshot: Towards Controllable Video Generation and Editing with Multimodal Conditions

TL;DR

Moonshot introduces a multimodal video generation model that allows control over appearance and geometry through image and text inputs, improving quality and versatility in video synthesis tasks.

Contribution

The paper presents Moonshot, a novel video diffusion model with multimodal conditioning and the ability to incorporate pre-trained image control modules without extra training.

Findings

Significant improvement in visual quality and temporal consistency.

Versatile multimodal conditioning enables diverse applications.

Easy integration with existing image control modules.

Abstract

Most existing video diffusion models (VDMs) are limited to mere text conditions. Thereby, they are usually lacking in control over visual appearance and geometry structure of the generated videos. This work presents Moonshot, a new video generation model that conditions simultaneously on multimodal inputs of image and text. The model builts upon a core module, called multimodal video block (MVB), which consists of conventional spatialtemporal layers for representing video features, and a decoupled cross-attention layer to address image and text inputs for appearance conditioning. In addition, we carefully design the model architecture such that it can optionally integrate with pre-trained image ControlNet modules for geometry visual conditions, without needing of extra training overhead as opposed to prior methods. Experiments show that with versatile multimodal conditioning mechanisms, Moonshot demonstrates significant improvement on visual quality and temporal consistency compared to existing models. In addition, the model can be easily repurposed for a variety of generative applications, such as personalized video generation, image animation and video editing, unveiling its potential to serve as a fundamental architecture for controllable video generation. Models will be made public on https://github.com/salesforce/LAVIS.