Adobe continues to actively implement artificial intelligence features. In addition to the capabilities of the Firefly Image 3 model in Photoshop, the company has also prepared something useful for video creators.
An experimental artificial intelligence-based function VideoGigaGAN can improve the quality of video by increasing image sharpness. It is noted that it can increase sharpness by 8 times without the usual artifacts such as flickering or distortion.
Adobe claims that VideoGigaGAN outperforms other Video Super Resolution (VSR) methods because it avoids common artifacts and flickering embedded by General Adversarial Networks (GAN). At the same time, it adds sharpness and detail - while most other systems cannot perform both of these tasks simultaneously.
The system copes well with improving the detail of skin texture, fine hair, feather details. These objects look quite realistic.
Previous VSR models struggled to generate rich details in the results. To address this issue, Adobe used "temporal attention" (reducing artifacts that accumulate over time), function duplication (adding details where they are missing), smoothing, and the HF shuttle mechanism to create the final result.
If VideoGigaGAN is added to programs like Premiere Pro or After Effects, it can make video editors' work easier, although using artificial intelligence to enhance people is a controversial practice.
There is currently no information on whether Adobe plans to implement this feature in real programs. The created clips are short and play back at only 12 frames per second. Developers will likely need to further refine this mechanism.
Source: Engadget
Comments (0)
There are no comments for now