What if a 3D scan were taken of the subject, and the artist then made the painting while viewing the 3D representation in some kind of augmented reality setup?
That would be 3D but still effectively painting from a "photo".
But then you're adding in several more processing layers away from reality: 1) the digital scan won't pick up the details sufficiently 2) the AR rendering will again strip a lot of detail
It'd also be incredibly difficult and straining for your eyes of focussing on a canvas and then on an AR image
It'd also be incredibly difficult and straining for your eyes of focussing on a canvas and then on an AR image