A variational autoencoder designed for video that compresses video frames into a latent representation for efficient processing.