Adobe Research has made a significant breakthrough with the introduction of VideoGigaGAN, a generative AI model that can upscale blurry videos by up to eight times their original resolution with incredible accuracy and detail. This groundbreaking technology, presented in a research paper, surpasses existing Video Super Resolution (VSR) methods by providing sharper and more lifelike results without introducing the common issues of flickering or other unwanted artifacts.
Generative Adversarial Networks (GANs) have proven effective in upscaling still images, but their application to videos has been challenging due to the flickering and distortions that can occur. VideoGigaGAN effectively addresses this problem, delivering the superior image quality of GAN models with the stability and smoothness of traditional upscaling techniques.
The research team at Adobe has released compelling demonstration videos that showcase the capabilities of VideoGigaGAN in full resolution. While some finer details, such as skin texture and creases, appear to be artificially enhanced, the overall results are surprisingly realistic and natural. This remarkable achievement has the potential to redefine the boundaries of video editing and production.
It’s important to note that VideoGigaGAN is currently only available as a research preview, and its future availability to consumers through Adobe’s Creative Cloud suite (including Premiere Pro) remains uncertain. However, the company’s previous exploration of diffusion-based upsampling, showcased at the MAX event in October 2023, demonstrates Adobe’s continued dedication to pushing the boundaries of video enhancement technology.
Adobe is not alone in this pursuit, with other industry giants such as Google and Meta also investing in the development of their own VSR upscaling solutions. As research and innovation continue in this field, we can anticipate further advancements that will empower creators and content producers with powerful tools to transform low-resolution footage into stunningly clear and detailed videos.
