Microsoft Research's Mirage gives video generation a persistent spatial memory that doesn't forget what's around the corner

Microsoft Research has introduced Mirage, a video world model that utilizes latent space to store scene information, significantly reducing compute time and graphics memory compared to traditional pixel-based point clouds. This approach maintains spatial consistency during extensive camera movements but currently struggles with tracking moving objects across segments. This advancement is crucial for practitioners in video generation, as it enhances efficiency and consistency in rendering dynamic environments.