Generative Video Compression with a Transformer-Based Discriminator