NOTE: This is not an entire hair shader - I'm showing you parts of a shader that do things in modular ways. You have to assemble these modules with other modules.

I'm still not at a finished hair shader here. I'm just showing you how much you can do in the shader alone. There's no geometry yet - this is a one-sided square on the left.

At this point I'd switch to all matmatic - no building in the mat room - too many nodes. What I'd want now is to layer another entire set of this to produce some hairs that are not strictly parallel to these hairs. I want them to slide and cross each other in interesting ways.

When I'm done, I predict I'll have around 100 nodes.

So this is where you guys were hiding all the interesting stuff. The "Anisotropic Node" thread sneakily turned into a procedural hair thread. You are so bookmarked! I guess I have no choice now but take some time off JCM-ing and make a hair prop or five. :biggrin:

I had totally run aground, Bill... thank you for these clues! Something new to chew on! The hair in its current implementation look reasonable from a distance, but unacceptable up close:

BTW, the dress is the GC version of Hongyu's Jersey Dress - default texture. I'll bet even he didn't know it was meant to be that colour.

Can't wait to plug this all in and make it work. I remember your tickler post on RDNA Node Cult and thought: 'there's a huge difference between that and what we're coming up with... something missing, here'.

And yes, it is more fun digging around and trying to get it, given a few clues. I think I've come up with a whole arsenal of new materials just from trying to find hair. Of course, mucking around in scripts is still that goal to shoot for: been reading your loom and ooooo lots to get the head around, yet. Need a primer on maths and nodes, but I probably be writing it, given all the experiments I'm doing. Must say, though, for at least my stuff I can actually see the basis for the material better in script form than in the mat room, believe it or not.
Anyway, I'll publish the more interesting code I come up with, and no more weird stuff, I promise. You'll be able to follow it just fine.

I'm so excited about this I'm even turning away chocolate!

bagginsbill:- I don't think i explained what i ment very well,lol..... i mean the hair eminating from a single point in the center of the shader and spreading outward towards the edges in a star type of effect... thankyou for the previous help... i must admit i thought "oh god hes going all math on me again",lol.... but i think im kinda understanding, and im gonna have a little play with what you have shown here....

i will post my shader soon too

so BB you do nothing fancy with the anisotropic node? you only do custom hair strands and just the anisotropic node without any trick for the highlights?

Quote - so BB you do nothing fancy with the anisotropic node? you only do custom hair strands and just the anisotropic node without any trick for the highlights?

Even I don't know what I do, yet. I'm still figuring hair out like you guys are. However, there are a couple more things I do. So far, nobody has complained of the issue that I believe requires attention and additional care. I don't like just telling you guys "here's 50 nodes, it does something important". Up to know, you don't even seem to be aware of the problem. And, I'm not convinced I have the solution, either. So I'd rather wait for you guys to catch up to me. You might think of something I didn't and if I lead you to my current half-solution, I may prevent you from finding a better one.

But first, you need to discover the problem.

RV found a problem already - the artifacts. But there are more problems with the anisotropic node.

Quote - bagginsbill:- I don't think i explained what i ment very well,lol..... i mean the hair eminating from a single point in the center of the shader and spreading outward towards the edges in a star type of effect... thankyou for the previous help... i must admit i thought "oh god hes going all math on me again",lol.... but i think im kinda understanding, and im gonna have a little play with what you have shown here....

i will post my shader soon too

OK I can do that too, but it's math. And then you'll probably say it isn't exactly a point, it's a short line, kind of a combination between the straight long part I did and the point part. Which would be a LOT more math.

I'm not really suggesting that hair has to be procedural - color maps are ok, especially for the scalp thingy. I just think it's good to know how to improve them with some procedural strands.

this is the problem that i found .
this is how the render looks

this is the UV
is it normal that the anisotropic goes in the middle? 

if i now the same model light above. so for example the light is 100% above this model. there is almsot no specular.is the anisotropic changing on the angle? 
so for example if you want to do dramatic lighing on top of your head then you dont get specualr on the hair. this is what i noticed.

aha plus sometimes the specular gets very bright. super bright

Super bright! Yes.

It depends so much on the UV map density. When the UV contour lines get closer together, the intensity increases to 100 times brighter in certain places. This is not acceptable.

i agree.

This is the way I prefer tackling this, Bill. Start at the beginning, and make sure core concepts are completely clear.

Quote - Consider the mathematical expression Abs(U - .5), i.e. the absolute value of the distance from the current U value to the fixed value .5. What does that do?

GMADone from the Node Cult defined u this way:
"I think of u (u_Texture_Coordinate, see image) as defining a line at the left of the texture (Zero or Black) with an even slope of positive 1 side to side so the result at the right of the texture is 1 or White, and a slope of zero up and down."

Know your enemy (or whatever). I can work with this description.

So, now I have sort of a point of reference for the expression (U - .5). The Abs bit is easy. Whatever U could be, subtract .5 and then turn the result into a positive (real?) number if it isn't already. So, per that definition, u is a range? (or: "even slope") of 0 to 1 from left to right, and a zero (0) range vertically.
I just have to nut that out thoroughly, or the rest of what you were doing isn't going to make any sense.

Maybe we should rename this thread "Anisotropics for Dummies".

Quote - Now consider procedural hair that forms horizontal hairs, using the V node to drive an fBm or Fractal_Sum. Suppose instead of using V to drive it, you used:
V + k * (Abs(U - .5))
where k is a small fraction such as .1 or .2. What does that do?

Right, so V works similarly to U, except the directions are turned 90 degrees. 0 to 1 vertical slope, 0 horizontal slope. And you're multiplying the sum of V + .1 and the absolute difference of U and .5.

Can U or V be represented as some numerical value, just so I can even begin to get a grip on this? Or will that actually prevent me from realizing the concept.

Major holes, here.

Quote - Super bright! Yes.

It depends so much on the UV map density. When the UV contour lines get closer together, the intensity increases to 100 times brighter in certain places. This is not acceptable.

I'm not completely certain if I understand correctly, Bill. I have that sort of highlight on straight hair where the UVmap has no converging textures. I naturally assumed it was the angle of viewing and that what I was seeing at the crown was the combined result of a much higher sum of shader-brightened geometry and anisotropic mapping (if I can call it that: what is determined by the U-Highlight_Size and v_Highlight_Size) than what I saw halfway down the face of the hair geometry (say, at ear level, where you looked at the texture sagitally.).

So if by "converging" and UVmap density" you meant "seeing the UVmapped texture at an angle or side on", then I agree.

Wow interesting thread. How did it ever end with those hair shaders? Did they move to a different thread?

And oh what a shame that all of Robynsveil's pictures aren't here anymore. It makes the thread sadly amputated...

(yes I know it's old! It just shows the dangers of hotlinking to external pictures in cases like these)

I suppose I could/should upload them again. I'm having a look at the fractal_sum node in PP2012 to see if the texture issues are still there. And yeah, you're right, TG: it makes for a clumsy thread without the piccies...

i went back to the anisotropic shader. and i just realized what was wrong. the node doesnt work with anisotropic textures.

in 3D the only way to control where anisotropic direction is with a special texture that you can make in photoshop.

this is for example for a CD or DVD.

this is for metal like here .

if we could use this kind of textures we could also use them for hair. we could control in which direction its going.

or can we use those textures?

     I didn't know about the hair and brushed metal applications.  I have been using the anisotropic node to get sparkly water.

large scale...

.

small scale...

.

Technically, water shouldn't have anisotropic highlights/reflections.  Anisotropic highlights come from fine grooves on a surface.  But it has a very cool effect on your sparkly water.  Which is to be expected - all your special effects are cool.

ice-boy- You're thinking about the problem and sort of missing a step.  The examples you're showing aren't useful maps for realistic anistotropic highlights, and are more like either a diffuse map with highlights and reflections burned in (first) or a bump map for scales (second).  The second doesn't even have a distribution around the center of the circles the way you seem to want.  Neither would be useful control maps for any kind of highlight.

Think about it like it's math.  U and V are just X and Y.  If you want a circle, you need the equation x ** 2 + y ** 2 = r ** 2.  That's a circle centered around 0, 0.  If you want that circle to shift from the center, then you need to add in the elements for setting the center.  IIRC , that's (x - x1) ** 2 + (y - y1)**2 = r ** 2, where the center is at x1, y1.  If you want the circles to repeat, then you need x1 and y1 to be a discrete (as in not continuous) function of x and y.  Something that will give you a series of numbers spaced apart by a function of r.  Sort of like a stair case function, where each step is r or r minus a bit high.

In general, if you want a map to control anisotropic highlights, or any other node for that matter, you have to consider what you're controlling.  Anisotropic has basically 3 parameters: intensity, U (x) scale, and V (y) scale (the other parameters don't seem to do anything).  So to control its distribution, you'd have to start mucking U and V scales.  Mind, those follow the UV map, not x and y in 3d space, so different mappings of the same mesh will have different effects.

A much simpler approach would be to stop thinking about how to mimic the effect of fine grooves and just make a map for fine grooves to add to your bump.  That said, it would probably be best to make one by hand, because maps like that metal one would just be totally off due to the burned in highlights and reflections.  Should be easy, though.  Just take random cloud in PS, use a radial blur filter on it, cut it into a circle, and then scale and duplicate it.