Data to be translated: Nanyang Technological University releases the Upscale-A-Video framework, addressing the quality issues in large-model video generation. This framework integrates functionalities such as super-resolution, denoising, and restoration, enhancing the quality of generated videos without the need for extensive training. It employs diffusion methods, combining local and global strategies to maintain temporal consistency. The temporal U-Net and recurrent latent code propagation modules effectively improve video quality, resulting in more natural and coherent outputs. It supports text prompts and noise level adjustments, increasing the diversity of generated results.