Something I have noticed is that Topaz Gigapixel struggles to upscale and add detail to pictures where an animal is the only subject. For myself, I have used low-res photos of a dog which an AI model should find easy to notice, and yet neither the standard models nor the recover model make a noticeable impact on the detail or smoothness of the image. That has to improve on a product like this. A good way to do this with pets would be to let the user upload multiple pictures of the same subject to inform the model.
On humans and faces it is simply exceptional, although it seems to overly smooth faces at high levels of recovery. It would be nice to see further options for the user to manipulate how the Face Recovery feature does its job: in terms of detail, sharpness, smoothness, or by adding multiple pictures of the same face to inform the model.
