I've heard recently about this company that makes consumer eye-tracking devices for laptops/tablets. Then I rembered that back in 80s I was reading about how military flight training simulators from that period (not sure if this still applies today) had eye-tracking, so that only what is in the focus was rendered with lot of details (as a test focus on just one word here and without defocusing try to precieve how small angle of focus actually is).

This made me start thinking - is it possible to make a game with this approach, and then implement tech similar to described so that only small area that is actually in our focus is vastly enhanced.

I've seen demo of the game that uses eye-tracking for control (no idea what came out of that at the end), so I'm guessing that tracking speed is not the problem...but not so sure about rendering part.

So, anyone with bit of knowledge (or just interest in topic), please chime in...

The speed your focus can move across a screen is far too fast for what I think you are saying to help in streaming graphical elements.

In open world games you get pop-up where things in the distance are not rendered in time, if you are saying that clearer, more realistic graphics would be possible if only a small section is rendered properly (ie the bit you look at... with outer areas using simpler models and textures) then yes that would be true.... but while eye tracking technology can easily figure out where you look on screen, the CPU/GPUs of today are nowhere near fast enough to stream new models and textures like that each time you move your focus.

The military tech you mentioned sounds more like a deliberate visual effect... the flight simulator will be rendering all of the screen as normal, then overlaying a blurring effect which follows your focus to make it seem more realistic. With surround screens it would be a cool thing to put into 3D video games, but would not release any graphical power (and I think would look really weird on any screen that didn't take up a big chunk of your field of vision)

to, i think, agree with highurtenflurst... i don't think they could track my eye fast enough to respond accordingly.

Yea you could use quality rings. The center circle is native res with full aa. Each ring out has lower res and worse aa. Having different models or texture quality would take up a lot more power than could be saved.

The problem is this would need to be locked 60fps or else quickly looking around would hurt your brain. Also the ring size would depend on screen size and distance from it. A smaller screen at a larger distance would need bigger sized rings. However this might be offset by the user having worse viewing conditions.

Thanks all for replies. 80s were so long ago, but if my memory serves me well, that military tech (I think it was flat shaded, Gouraud at best) was doing more or less what we're discussing - not sure if there were quality rings, or just one wider with everything outside of it rendered with lower details, but, if IIRC, they were tracking eyes really fast...

Of course, this is just speculating, I'm not sure this is even achievable with current engines - I'm guessing some sort of adaptive tessellation and partially resident textures approach might do the trick, though I'm more interested to learn if this is achievable with some sort of fully voxel based engine in the future.