W.A.L.T

W.A.L.T is a transformer-based model developed by researchers from Stanford University, Google Research, and the Georgia Institute of Technology. It employs a causal encoder to jointly compress images and videos into a unified latent space, enabling training and generation across modalities. The model utilizes a window attention architecture tailored for joint spatial and spatiotemporal generative modeling, achieving state-of-the-art performance on established video and image…