Good article Mark. I’ll also point out that some of the techniques used by fast renders just don’t look right in dual rendered 3D. As an example normal maps and bump maps are used to give an otherwise flat surface more apparent surface detail without adding polygons. While these widely used techniques look great in 2D the effect falls apart in 3D. Because your brain prioritizes the 3D effect it gets via 2 separate images, the effect used by normal/bump maps goes to looking odd/flat. You can still use these techniques at an apparent distance beyond the 10–12 meters where human vergence depth perception is useless.
The only choice you are left with is to add polygons to make up the detail and that of course slows down the overall rendering. There is going to be a limit of how many polygons you can dual render at 60–90 FPS. This ends up being a design decision, do you go with more polygons per object with a limited number of objects, or less polygons and more objects. I think for the next few years studios are going to need to favor stylized objects, like what you see alchemy labs does, over hyper-realism.
