www.analyticsdrift.com
Image source: Analytics Drift
Adobe launched a paper on April 18th on a new generative AI, VideoGigaGAN, which upscales low-quality videos by 8x their original value.
Image source: Adobe
VideoGigaGAN is built upon the large-scale image upscale GigaGAN and has significantly improved the temporal consistency of upsampled videos by adding temporal modules.
Image source: Adobe
VideoGigaGAN aims to provide the best of both worlds—the higher image/video quality of GAN models with few flickering or distortion issues across output frames.
Image source: Canva
They introduced a “flow-guided propagation module” and a “high-frequency features shuttle” for sudden drops in detail to ensure consistency among a video's frames.
Image source: Canva
To show its potential, they have given a demo where an old movie from the 1940s goes from running at a 480x360 image resolution to a crisp 1280x960.
Image source: Adobe
In the past year, the company has been focusing heavily on implementing AI into its software from the launch of Firefly and Acrobat’s new assistant.
Image source: Adobe
For now, VideoGigaGAN is just a research preview, and there are no immediate plans to bring it to Premiere Pro or its Creative Cloud Suite.
Image source: Canva
Produced by: Analytics Drift Designed by: Prathamesh