Photorealistic Video Generation with Diffusion Models

arXiv Project

We present W.A.L.T, a transformer-based approach for photorealistic video generation via diffusion modeling. Our approach has two key design decisions. First, we use a causal encoder to jointly compress images and videos within a unified latent space, enabling training and generation across modalities. Second, for memory and training efficiency, we use a window attention architecture tailored for joint spatial and spatiotemporal generative modeling. Taken together these design decisions enable us to achieve state-of-the-art performance on established video (UCF-101 and Kinetics-600) and image (ImageNet) generation benchmarks without using classifier free guidance. Finally, we also train a cascade of three models for the task of text-to-video generation consisting of a base latent video diffusion model, and two video super-resolution diffusion models to generate videos of 512x896 resolution at 8 frames per second.

@article{gupta2023photorealistic,
  title={Photorealistic Video Generation with Diffusion Models},
  author={Gupta, Agrim and Yu, Lijun and Sohn, Kihyuk and Gu, Xiuye and Hahn, Meera and Fei-Fei, Li and Essa, Irfan and Jiang, Lu and Lezama, Jos{\'e}},
  journal={arXiv:2312.06662},
  year={2023}
}