Skip to content

VideoGigaGAN

Upscale video with high-frequency detail and temporal consistency

Visit official site

What VideoGigaGAN is

The model is described as a video super-resolution model built on GigaGAN, a large-scale image upsampler, that produces video with high-frequency detail and temporal consistency. It adds temporal attention layers and a feature propagation module and uses anti-aliasing blocks to reduce aliasing effects.

What you can do with it

  • Perform video super-resolution
  • Keep temporal consistency across frames
  • Generate high-frequency detail in the output
  • Apply anti-aliasing processing to the result

Application scenarios

  • Raise low-resolution surveillance video to high definition to identify details
  • Upscale old films or older video in editing to improve the viewing experience
  • Upscale web video for large-screen display or high-definition printing

Model notes

  • The model is based on the GigaGAN large-scale image upsampler
  • Temporal attention layers and a feature propagation module improve temporal consistency
  • Anti-aliasing blocks reduce aliasing effects
  • Comparisons on public datasets show 8x super-resolution video results

Pros & cons

✓ What we like

  • Anti-aliasing blocks reduce aliasing effects
  • 8x super-resolution video results are shown on public datasets

! What to watch out for

  • The approach starts from a large-scale image upsampler
  • Temporal consistency depends on the added attention and propagation modules

FAQ

What is it built on?

The material describes a video super-resolution model based on GigaGAN, a large-scale image upsampler.

What result scale is shown?

The material states 8x super-resolution video results were presented on public datasets.

Last reviewed: 2026-11-15

More AI video generation tools

View all →

How we review