Skip to content

STAR

Real-world video super-resolution with diffusion

Visit official site

What STAR is

The framework combines a text-to-video diffusion model with video super-resolution to restore detail and keep temporal consistency. It was developed by Nanjing University and ByteDance among others.

What you can do with it

  • Enhance spatio-temporal quality with diffusion
  • Reduce artifacts from complex degradation
  • Apply a dynamic frequency loss

How to start

  • Download code and resources from the project site
  • Prepare a low-resolution video file
  • Configure the environment and run the model

Pros & cons

✓ What we like

  • Multiple real-world video sources are supported
  • Bilibili and VideoLQ videos are named

! What to watch out for

  • Processing needs a low-resolution video as input
  • The guide requires configuring environment and parameters

FAQ

What problem is addressed?

The material describes over-smoothing in traditional GAN methods.

Which sources are named?

The material lists Bilibili and VideoLQ among video sources.

Last reviewed: 2026-11-15

More AI video generation tools

View all →

How we review