Adobe continues to actively introduce artificial intelligence features. In addition to capabilities of the Firefly Image 3 model in Photoshop The company has also prepared something useful for video creators.
An experimental AI-based feature called VideoGigaGAN is capable of improving video quality by sharpening the image. It is noted that it can increase the sharpness by 8 times without the usual artifacts such as flickering or deformation.
According to Adobe, VideoGigaGAN outperforms other VSR (Video Super Resolution) sharpening methods because it avoids the common artifacts and flicker introduced by GANs (General Adversarial Networks). At the same time, it adds sharpness and detail — whereas most other systems cannot do both at the same time.
The system does a good job of improving the detail of skin texture, fine hair, and feather details. These objects look quite realistic.
Previous VSR models had difficulty generating rich details in the results. To fix this problem, Adobe therefore used «time attention» (reducing artifacts that accumulate over time), feature propagation (adding details where none exist), anti-aliasing, and the HF shuttle mechanism to create the final result.
Adding VideoGigaGAN to programs such as Premiere Pro or After Effects can make video producers’ jobs easier, although using artificial intelligence to enhance humans is a controversial practice.
There is no information yet on whether Adobe plans to implement this feature in real programs. Currently, the clips created are short and play only at 12 frames per second. Probably, developers will have to improve this mechanism.
Source: Engadget