Apr 25, 2024 13:00:00

Adobe's research team develops AI 'VideoGigaGAN' that can upscale video while restoring details

Adobe has announced ' VideoGigaGAN ', an AI capable of upscaling video. Compared to existing video upscaling AI, VideoGigaGAN can produce clearer images and also reduces flicker between frames.

VideoGigaGAN

https://videogigagan.github.io/

As an example of the power of VideoGigaGAN, Adobe has released a demo in which a 128 x 128 pixel video is upscaled to 1024 x 1024 pixels, eight times larger in both width and height.

Below is a side-by-side comparison of the original image (left) and the upscaled image (right) with the same size. You can see that the image is upscaled to a fairly high quality.

When the waterfall image is upscaled, it looks like this. The upscaled image (right) has a clearer depiction of the water flow and plants compared to the original image (left).

According to Adobe, VideoGigaGAN is superior to existing video upscaling AI in terms of reproducing details and reducing flickering. Adobe is highlighting the high performance of VideoGigaGAN by comparing the processing results of video upscaling AIs such as

BasicVSR (top left), BasicVSR++ (top right), and TTVSR (bottom left) with those of VideoGigaGAN (bottom right).

For more technical details about VideoGigaGAN, please see the following link.

[2404.12388] VideoGigaGAN: Towards Detail-rich Video Super-Resolution
https://arxiv.org/abs/2404.12388

◆ Forum is currently open
A forum related to this article has been set up on the official GIGAZINE Discord server . Anyone can post freely, so please feel free to comment! If you do not have a Discord account, please refer to the account creation procedure article to create an account!

• Discord | 'Do you have a video you want to upscale? What kind of video is it?' | GIGAZINE
https://discord.com/channels/1037961069903216680/1232985589129809981