Next Gen Eyes

Spiralface

So I'm curious as to how artists can approach eyes on "bleeding edge" engines like Unreal or Marmoset that get you Super fine results.

Are we at the point where we are making 2 layers of geo objects for the eyes. (One for the cornea, and one for the iris / sclera) Or are we still sticking with a single geo object painted?


I know for most "low poly" production, you keep it as one object, but I'm curious if you still end up doing this for high end graphics production, like if you where to make something for Unreal engine.

odium

I find that eyes works amazingly well if you add an envmap effect to them and invert the cornea on the normal map. The reflection really adds depth to them.

Shaffer

That invert the cornea type stuff is what I have been contemplating.

Snader

I wouldn't be surprised if we're actually at the point of doing this:
making a single geo, and then using shader tricks (such as parralax/offset bump maps) to make the dent of the iris and such.

Having such a map would give a pretty simple way to increase small detail such as clothing folds, skin wrinkles, and whatnot.

Having it all as a single extra texture/drawcall also sounds more simple to turn off for lower performance platforms, but I'm not entirely sure about that.

Personally, I'm still working at the point of "only using 500 triangles so I'll paint the eyes on the face"

cman2k

I made this eye shader a while ago for 3dsmax. I believe it is pretty similar to some of the high-end shaders I have seen in games like Mass Effect 2 and Left4Dead 2, as I tried to emulate many of their features.

This uses a single layer and shader tricks, as mentioned above. Using two actual layers of geo is still mostly unpractical because to get the most out of that you'd need actual raytraced reflection/refraction, I think.

Spiralface

cman2k wrote: »

I made this eye shader a while ago for 3dsmax. I believe it is pretty similar to some of the high-end shaders I have seen in games like Mass Effect 2 and Left4Dead 2, as I tried to emulate many of their features.

This uses a single layer and shader tricks, as mentioned above. Using two actual layers of geo is still mostly unpractical because to get the most out of that you'd need actual raytraced reflection/refraction, I think.

I was thinking less along the lines of raytraced refractions, and more along the lines of having things like the specular not sample the normal maps.

Currently the issue I keep on thinking of is having a normal map for your iris and slight veins but if I just flat out apply that, then your reflections and specular would also sample off of the Normal map geo.

By separating the geo, I would be able to have the eye ball with the iris have a normal map and some subtle reflections, but then have a completely translucent shader on the "Cornea" object that only picks up high gloss specular and reflections. I like the eye shader though, Is there a way that you got the Cornea and cube map reflections to not be affected by the underlying normal map?

cman2k

yeah in HLSL that is pretty simple to do compared to a set pipeline like UDK. When I calculate the specular I use one set of normals, then I do the parallax and second normal seperately for the inner iris, then I add them together.

That stuff might be possible in UDK, but it'd probrably be more complicated? I imagine you'd be doing a lot of masking and stuff...? I am not experienced enough in UDK to say, but you are correct in that using a secondary piece of geo with a seperate shader could potentially be a simpler hack-around.