It’s a time-consistent diffusion model of video super-resolution and works awesome, see how clear those lines become!
More details are below ⬇️ ⬇️
The noise level is adjustable and can be trade-off between recovery and generation
And hints can be used to guide texture creation
That’s all, the original author @alexcarliera
Project Homepage:
shangchenzhou.com/projects/upscale-a-video
GitHub repository:
github.com/sczhou/Upscale-A-Video