VidTwin: Video VAE with Decoupled Structure and Dynamics

TL;DR

VidTwin introduces a novel video autoencoder that decouples content and motion into separate latent spaces, enabling high compression, quality reconstruction, and scalable video generation.

Contribution

It proposes a compact, explainable Video VAE with decoupled structure and dynamics latent spaces, enhancing video compression and generation capabilities.

Findings

Achieves 0.20% compression rate with PSNR of 28.14 on MCL-JCV dataset

Demonstrates effective downstream video generation tasks

Shows scalability and explainability in latent video representation

Abstract

Recent advancements in video autoencoders (Video AEs) have significantly improved the quality and efficiency of video generation. In this paper, we propose a novel and compact video autoencoder, VidTwin, that decouples video into two distinct latent spaces: Structure latent vectors, which capture overall content and global movement, and Dynamics latent vectors, which represent fine-grained details and rapid movements. Specifically, our approach leverages an Encoder-Decoder backbone, augmented with two submodules for extracting these latent spaces, respectively. The first submodule employs a Q-Former to extract low-frequency motion trends, followed by downsampling blocks to remove redundant content details. The second averages the latent vectors along the spatial dimension to capture rapid motion. Extensive experiments show that VidTwin achieves a high compression rate of 0.20% with high reconstruction quality (PSNR of 28.14 on the MCL-JCV dataset), and performs efficiently and effectively in downstream generative tasks. Moreover, our model demonstrates explainability and scalability, paving the way for future research in video latent representation and generation. Check our project page for more details: https://vidtwin.github.io/.