Adobe is on the cutting edge of technology with their latest project which involves developing an artificial intelligence technology capable of enhancing low-resolution video to 8 times its original quality. The project, known as VideoGigaGAN, is a generative AI model that upscales low-quality videos, preserving important details while significantly increasing their resolution.
The Power of VideoGigaGAN
The innovative VideoGigaGAN AI model has the ability to upscale videos by up to eight times their original resolution without sacrificing stability or crucial aspects of the source material. The project showcases demo clips on its website, demonstrating its impressive capabilities. For instance, it can transform a blurry 128×128 pixel resolution video into footage running at a resolution of 1,024×1,024 pixels, showcasing remarkable enhancements in quality.
The Intricacies of the Technology
Utilizing the power of a generative adversarial network (GAN), VideoGigaGAN is able to create realistic images. Adobe’s version is specifically based on GigaGAN, specializing in upscaling generated content and real photos. To address challenges such as artifacts in the upscaled videos, Adobe researchers implemented various techniques including a “flow-guided propagation module” for consistency among frames, anti-aliasing to reduce artifacts, and a “high-frequency feature shuttle” to compensate for drops in detail.
Potential Integration into Adobe Products
Considering Adobe’s recent focus on integrating artificial intelligence into their software, it is highly likely that VideoGigaGAN may find its way into an upcoming product or as a standalone app. Recent previews at events like Adobe MAX 2023 suggest that this technology could be incorporated into products like Project Res Up, offering users the ability to upscale old videos and low-quality footage significantly. While the exact release date is uncertain, the potential of VideoGigaGAN in revolutionizing video enhancement is promising.
You might also like
As VideoGigaGAN continues to undergo development, challenges such as processing limitations and object rendering issues still need to be addressed. The potential for this AI technology to reshape the way we enhance videos is immense, and the industry eagerly anticipates the impact it will have on the future of video editing and enhancement.
