Metal Specular Maps - Your Methods
polycounter lvl 18
I've come to the conclusion that all my specualr maps pretty much look the same. I've tried a few different things but they always seem to look a bit naff.
What do you find works best as a specular map ATM for metals? Obviously metals are shiney but at the same time, dull.
ATM, I usually make it quite dark, with bright highlights for edges and and over all with a tint of blue.
What are your methods? (With examples hopefully!)
What do you find works best as a specular map ATM for metals? Obviously metals are shiney but at the same time, dull.
ATM, I usually make it quite dark, with bright highlights for edges and and over all with a tint of blue.
What are your methods? (With examples hopefully!)
Replies
In an engine you can use a double spec (works well on hair too). One highlight is tight, and the other spread out.
Diffuse:
http://www.quake2evolved.com/odium/wall1_d.jpg
Local:
http://www.quake2evolved.com/odium/wall1_local.jpg
Spec:
http://www.quake2evolved.com/odium/wall1_s.jpg
In a game engine or in a 3d package for rendering?
In an engine you can use a double spec (works well on hair too). One highlight is tight, and the other spread out.
[/ QUOTE ]
aw man! i thought i was being clever!
there's a name for it and everything? fack, u guys r 2 good 4 me!
seeing other game artists use a tight spec always bugged the shit out of me because it sacrificed surface detail for a shitty little spec. i usually made it a dull glossiness and let the spec map do most of the work. i wanted a dual-spec shader, but unfortunately due to a number of circumstances i could never get it.
here's an old R&D pic to better explain the idea:
so nice of an effect. i'd love to play with tech like it that displayed on a normal map! if anyone knows of an engine or max fx that supports it, i'd love to play with!
-kp
I usually go completley loose with the gloss mask and make it look some-what wavy - as if the specular pass was being refracted. From there I just tone the spec and environmentmap mask.
[ QUOTE ]
In a game engine or in a 3d package for rendering?
In an engine you can use a double spec (works well on hair too). One highlight is tight, and the other spread out.
[/ QUOTE ]
aw man! i thought i was being clever!
there's a name for it and everything? fack, u guys r 2 good 4 me!
seeing other game artists use a tight spec always bugged the shit out of me because it sacrificed surface detail for a shitty little spec. i usually made it a dull glossiness and let the spec map do most of the work. i wanted a dual-spec shader, but unfortunately due to a number of circumstances i could never get it.
here's an old R&D pic to better explain the idea:
so nice of an effect. i'd love to play with tech like it that displayed on a normal map! if anyone knows of an engine or max fx that supports it, i'd love to play with!
-kp
[/ QUOTE ]
Or just use HDR imaged based lighting! The lighting setup we use right now works basically like this: You get your main bright highlight from the sun, and then you get more ambient highlights from the rest of the sky. Its perfect for metal.
This double-spec technique intrigues me. I would like to see more examples if anyone has them. I assume you guys are using the same spec map for both highlights and just blending between a broad and a tight highlight? Couldn't you fake this by using a different highlight falloff curve ?
Once you have your src, it can be applied either like reflection or like diffuse. If reflection, the shader constantly factors in your viewing angle, so the reflection moves across the surface as your viewing angle moves. Like specular.
There is a shader technique I've seen for masking different mips of the same reflection src, so you could have smooth sharp-shiny metal and buffed blurry-shine metal per-pixel-masked on the same surface. Heard that's costly though, anyone using this? Or you could simply use two materials instead, one using a lower mip than the other. Wish I had pics here. We have support for this, but all the wrinkles haven't been ironed out yet.
If applying it like diffuse, then the view angle is not factored in, so it works more like bounce lighting. Then you want to use a very soft blurry source. ATI's cubemap tool is great for this, if you're using cubemaps. PRT does an even better job, since it isn't really pixels.
Multi-specular is pretty cool, does exactly what you see in killingpeople's image. We support it, works nice. Base material has everything (diffuse, emissive, bump, etc.) plus a wide specular power. Then you add another one on top, tighter power. Both speculars work well when subtle, since it's an add, blows out to white really fast. I'll try to rustle up some pics.
looking forward to what those look like eric.
here are some more of those tests i made:
Here's a quickie example.
I tried to whip up a nice HDR cube for comparison, but no time. I agree with EQ, HDR cube gives an accurate specular since reflect and specular are both in the end emulating the same thing.
A cube or PRT gives you all the scene colors, (yellow sun, blue sky) whereas each specular is a single color and only a radial gradient at that. HDR tosses intensity ramping into the mix, so you get more contrast, only the brightest reflect. Looks less chromey.
Downside to cubes/PRT is they emulate always-distant semi-fixed lighting. Engine has to recreate them at intervals if lights move (comes at a cost). Specular works great for moving lights... like a character walking down a hall, or wielding a flamethrower, or both.
looks AWESOME eric. i want to play with that! i wonder, with a setup like that, is there any reason to have two different speculars? can't you achieve the same effect with one gradient? also, can you impliment the use of a color spec for one of the spec inputs or somewhere in the pipeline?
Specular isn't really a colorable gradient, you can't set different colors for the center vs. the rim of the highlight. You can control per-pixel how that highlight changes color, but not color that conforms to the shape of the highlight itself. Not sure if that makes sense. If someone makes a shader for you it'll become pretty obvious pretty quick I think.
EQ, in the bit above where you referenced having a great cubemap, were you talking about an environment map for reflections or an HDR map for image based lighting?
[/ QUOTE ]
HDR map for imaged based lighting is best, but a cubemap can help too.
[edit] Model is gay, balls are touching.
I 1up earthquake's scratchy balls with my mighty science. Behold brushed metal!
It uses my high dynamic range image based anisotropic lighting shader model renderer 2.05b. There is a tangent-space direction map (much like a normal map but pointing along the surface) on the model that determines the direction of the grooves or brush strokes on the surface. The shader uses the direction map to "stretch out" the diffuse and specular hilights perpendicular to the grooves/brush strokes.
Here is a difference between phong and anisotropic lighting with a more extreme case on the right:
It uses image based lighting so here it is in different environments:
There is also a blurred noise specular map on it to illustrate which way the brush strokes are oriented on the model. In the direction map I use all the brush strokes are horizontal, so the blurred noise just happens to match up. This is the texture:
And finally, because it's kind of hard to show off totally sweet metal in still shots, a video (compressed with xvid):
http://www.reinot.com/8monkey/videos/aniso-cybork.avi
That anisotropic business looks saweet , got to study that further ! P.s - EQ ahaha !
We actually have a program that lets you paint on it to "comb" the surface, i'de show some examples but its all NDA stuff, sorry.
[/ QUOTE ]
So do you have an example then where the direction does change direction? This one looks like it's just following the blurred black and white noise around a horizontal axis with no actual changes based on the surface direction itself.
two years ago I did some stuff for university where I generated such per pixel rotation based on vertex normals. the directions were used for rotating "screening" patterns.
(scroll way down)
http://crazybutcher.cottages.polycount.com/index.html?/uni/npr/assign2.html
painting such a rotation map by hand would definetely require a tool that gives you a "direction" brush, which then outputs things as RGB, similar to painting "normals" just that they would be normalized and have a Z of 0, ie vectors that lie in the tangentspace plane.
such a direction texture would likely look similar to a normal map, I think. as it would be like a rotation matrix for each pixel. as 2d rotation would be enough R/G could store sin/cos of the orientation and maybe use B as stretching factor..
two years ago I did some stuff for university where I generated such per pixel rotation based on vertex normals. the directions were used for rotating "screening" patterns.
(scroll way down)
http://crazybutcher.cottages.polycount.com/index.html?/uni/npr/assign2.html
painting such a rotation map by hand would definetely require a tool that gives you a "direction" brush, which then outputs things as RGB, similar to painting "normals" just that they would be normalized and have a Z of 0, ie vectors that lie in the tangentspace plane.
[/ QUOTE ]
Hmm, now that you mention it, there is a way to comb normals built into max (edit normal modifier) I wonder if there is any way to bake that out into a map in a stored RGB format.
You're right, it is just a horizontal map, I was too lazy to comb it in Groom. The map just stores a vector actually, the grooves are infinitely long. When scaled and biased, it ends up looking like this:
[/ QUOTE ]
Awesome, I appreciate the show/tell.
Hey CB that's pretty cool, I like the hatching on the globe.