The one thing that Mask AI does well is that it applies the mask and blur effect intelligently, which means that it works well when you have delicate hair-strands you need to leave in focus and blur the background. So it would need to update its algorithm to simulate lens blur which is far more complex algorithm and it needs a lot of manual control to be able to control various aspects of the effect. That being said, the second problem of Mask AI is that it blurs the background with Gaussian blur algorithm that looks nothing like real thing. You might as well do it in Photoshop in multiple passes, because Mask AI is surly going to crash, and combine it as one depth mask. Soon as you have something more complex than that, with multiple objects in foreground, middle ground and background coming in and out of focus, gradually, than you need far more complex masking. If only thing you need to do is blur background and leave a person shot from waist up in focus, than its pretty straight forward. First and most obvious problem is the complexity of depth of field in some images and relative simplicity in other images. Well there are few problems with that, which would be something Topaz would need to improve if you are trying to simulate shallow depth of failed.
0 kommentar(er)
