Attached Link: Node Based Human Irises at the Node Cult

Follow the link.

Woo Hoo! Thanks Bb!!

Looks fantastic, thanks ! :)

WOOT! bB

Thanks a lot BB...

Can't wait to try it out.

JB

This is great!  I was just looking for something like this, last week.  I ended up re-mapping the irises, instead.

One problem, though.  Poser 5 doesn't have a Wave_2D node option.  Is there any way to wrangle the same effect out of P5, using a math function with sine, or something?

Not to that level of detail or complexity necessarily, but some approximation of that layout would be super.

i agree with pjz, want those striations. Hey BB, just give us a hint how to get started...we should try to learn how to do this stuff.  What would be the approach to getting complexity in the iris?

::: Opera :::

It would also be good if the pupil edge was a little softer (edge blend node? )

bookmarking.

An interesting technique.

Attached Link: Cartesian to Polar Coordinates - Material Room

I can do those radial fibers, no problem. Just have to use polar coordinates - very easy. If you want to race me, I'd love that.

The attached image is my brick shader which was designed for rectangular coordinates, tricked into making radial bricks by changing U and V into R and Theta.

Follow the link for a fascinating discussion of the use of polar coordinates and the importance of accuracy in the arctan function emulation. You need arctan to do polar and Poser math nodes don't have it. So I use an approximation. My buddy byRo used a different one. He thought his was good enough as his error was less than 1 part in a thousand. But that's not good enough.

Technical? You bet - dig in and you will be a node warrior too!

Cage - I didn't realize P5 had no Wave2D. No worries, the polar coordinates approach won't need it.

Nah I am way behind you on this.  The best idea I could come up with for how to do this involved using a bitmap as a seed, which kind of sucks the cool out of the idea.  I'm not a math head at all, I'm sure there is some application of basic math that describes a radial pattern but I have no idea what it would be. 
edit: but I can learn! :)

bagginsbill, have you checked out Jack Crenshaw's book "Math Toolkit for Real-Time Programming"?  This is really a dissertation on decades' long experience crunching complex mathematical routines into real-time embedded systems.  ArcTangent is covered in its own chapter. ;)

Now, just because this deals with embedded systems doesn't mean that something like old game approaches with limited angles and sin/cos tables.  Here, accuracy is just as important as well as error minimization.  This would need to be coded, unfortunately.  In nodes, it might be quite a few - a few constants (using tan() on two), abs(), and a significant amount of math for the nodes.  If there is a function node, this could reduce the overall node count.

Matmatic takes your formulas and implements them in nodes.

So for example when I type Bump = .03 * Bias(100 * (U + V) % pi, .7), it figures out that you can do that with 5 nodes and generates the shader you see in the image above.

So I don't really care any more that there is no formula node.

And I already have an extremely accurate and cheap conversion to polar coordinates. It is this:

So I don't really need any reference material for this purpose - the issue has already been resolved and tested extensively. With the above code included in a script, the U and V coordinates are now polar instead of rectangular. The rest of your script can then deal with U and V as usual and whatever you've described will be spun around a circle.

Bagginsbill's handy function to make conditional evaluations look more readable

def IF(test, a, b):
 return Blend(b, a, test)

A limited arcsin function that returns between 0 and pi/2

ONLY WORKS for x = 0..1

def limitedArcSin(x):
 a0=1.5707288
 a1=-0.2121144
 a2=0.0742610
 a3=-0.0187293
 x2 = x * x
 x3 = x2 * x
 return (pi/2 - sqrt(1 - x) * (a0 + a1x + a2x2+ a3*x3))

Bagginbill's extrapolation to get theta -pi < theta < pi

def getTheta(x, y, r = None):
 if r is None:
  r = sqrt(x * x + y * y)
 sinval = Abs(y)/r
 theta = limitedArcSin(sinval)
 return IF(x <= 0,
   IF(y <= 0, -pi+theta, pi-theta),
   IF(y <= 0, -theta, theta))

Bagginsbill's class to convert to polar coordinates because

we can share the sqrt part for r and theta.

class PolarCoord:
 def init(self, x, y):
  self.x = x
  self.y = y
  self.r = sqrt(x * x + y * y)
  self.theta = getTheta(x, y, self.r)
pc = PolarCoord(2 * U - 1, 2 * V - 1)
U = pc.theta
V = pc.r

fascinating

Very interesting, but now need to figure out what to use this type of material room wizardry on.  Hmmmm. 

bagginsbill, looks like you are doing a partial series for ArcSin().  I take it that the accuracy-error ratio is pretty good? :)

Yes - that partial series is accurate to within 1 part in 50000, worst case, and most of the time it's correct to better than that.

Wow my head hurts! Too much math! LOL! Can't wait to see where it takes us though! All this because I asked for Iris nodes over at RDNA? Wow!

Here's where I'm at so far. I have the polar coordinates thing working. I have a version of the Fractal_Sum node that produces radial noise. I can make layers of those tiny twisty fibers, although I haven't got enough control over them yet. And I can get some of that smoky ring stuff going. 

The WIP render shows how things look now. There are 5 layers of effects, 3 fibers, and 2 smoky. It took 155 nodes to do this. 

I need to refine how the effects are controlled and implemented, and I need more effects. Making the effects is hard. Once I have a nice collection of effects that you can layer on top of each other, you should be able to do some pretty complex stuff with a few lines of script.

I like it. I think this is going to be a really useful shader. It would be so hard to draw this by hand.

Man! And here I thought I new a good bit about the material room shaders! NOT!
So, am I correct in thinking that using this procedure instead of image based textures will speed up render times? And better yet, improve the workflow for animation?

That's a good start there, bb!  If you can manage more of the fibrus look, it may be the eye replacement we all want! :)

Can't wait to see the jaw-dropping finale.

Procedurals don't incur the "loading texture" time that big images chew up, so there is that savings. But in general, extremely complex procedures like this that have to use a lot of "noise" nodes (not one or two, but eight or ten) are actually somewhat slower than using an image. However, unless you're going to zoom the iris to 1000 pixels, you're not going to notice any speed difference. 

I can't see it providing any benefit for animation, unless you want to actually animate the texture itself, which is certainly possible. Many cool sci-fi effects could be done by altering the parameters to the shader in an animation.

Kurome - thanks. Yes I must figure out how to get the fibers to do more complex things - clumping, and more individual twisting. There are over 720 fibers in that image. Getting each of them to do its own thing without giving it it's own node(s) is a mind bender.

That is looking really, really good.  I think you ought to consider marketing that, because it kicks all kinds of butt.  Getting reasonably good eye materials is a chore, people always tend to leave in goofy reflections baked into the texture and this is EXTREMELY flexible and cool.  Really excellent achievement, I expect people will be ripping this off for years to come.

To me the big benefit is not so much speed, but flexibility - from what I can see, this technique can be tweaked pretty easily and quickly to come up with a very large variety of eye textures that will react to light the way they should, instead of having to painstakingly paint out flaws or add in variations with a paint program.

The most interesting thread since I-don't-know-when.
I'm a non-math person...  I appreciate the purity of mathematics as language, but I never learned to speak more than a few useful phrases sorta thing.

Are we talking about using a blend node over your existing shader types, or is this going to be a completely new hader?

Yes and one other thing is neat - the pupil radius and iris radius are parameters. This means you can design an iris, and then at the push of a button it generates shaders using that design for every character you have at once. I just need to get the coordinates for them all into the script. I already have V3, V4, and Jessi coordinates in it. You can add more and it's not hard to figure out. I made a testing shader where you give it an existing texture map and a guess as to what parameters to use. It then renders half with the procedural and half with the image file. You can see where they don't line up and adjust the parameter. Once you dial in the right number, its good forever.

Slinger this is completely new. I started with a blank sheet of paper, so to speak. Because I'm exploring the solution space, my script is currently very ugly with all kinds of false starts and bad choices in it commented out.

Once I determine the fundamental sets of functions you'll want to use, I'll refactor it and it will be all clean and simple to use.

Quote - ...Once I determine the fundamental sets of functions you'll want to use, I'll refactor it and it will be all clean and simple to use.

Gawd bless ya. ~lol~

Does this work with the existing Iris/Pupil material zones, i.e. does the IrisDilate morph slider still do what it's expected to do?  I'd expect it does, simply because the zones are fixed regardless of what texture/shader is applied to them, but I thought I should ask.

pj: yes it works with morphs. For future reference, a procedural texture, not just this iris one, will behave just the same as an image based one. An image is nothing more than a two-dimensional math function that is driven by (called with) U and V coordinates. So, too, is my iris shader driven by U and V coordinates. When you morph a prop, the U and V coordinates move, thus morphing the texture automatically no matter what kind it is.

By the way, there's nothing stopping you from running the procedural shader to produce an image which you save and use from then on. 

Does everybody understand that? It's called texture "baking", although true texture baking involves more than that.

Basically, you install the eye shader on a one-sided square, using only the Alternate_Diffuse channel so that the current light settings have no influence on the render. This is how I made the image I posted earlier - it is just the procedural applied to a one-sided square, rendered, then saved and uploaded.

An image like that (one with the whole eye including the whites) can be used directly just the same as all the hand-made ones.

As a merchant, you could generate several thousand of these in a few hours, no two alike, all photorealistic, and sell the whole lot for $5 because it suddenly is no work at all.

Oh, wait a minute. 

I just realized two things:

1. I should find out how much merchants make selling eye textures before I decide what I'm giving away for free. Perhaps I don't realize the total market size? I've often wondered about that. I'm not going to change my mind if the total market for eye textures is only $500 a year. But if I could realistically expect to make $10K or more, maybe I ought to rethink this. Any opinions, people?

2. If I do give it away, I'm going to make a lot of enemies as I undercut the market, right? Opinions?

Yes, I think if you made some move to make this proprietary and market it, you could make a great deal of money.  Considering that a large moderate quality eye package typically sells for $5-15, putting this kind of power and flexibility in somebody's hands could very easily go for a similar or even somewhat higher price.  Note that common brokerage agreements give the broker 50%, which means $2.50 - $7.50 per unit for you.  I'd give you $20 for this for absolute certain.  

If on the other hand you decided to give it away free, you would be up for sainthood right along side Anton Kisiel.  Matmatic is certainly cool, but this technique has a concrete application in every single human figure render (which is basically all of Poserdom).  

Some kind of dialog interface for basic layout would probably be required but I expect that is the easiest part for you (it sure would be for me).  Coordinates for a variety of figures would be very desirable as well.
http://market.renderosity.com/mod/bcs/index.php?Start=1&SearchTerm=eyes many examples

Note that nobody has a one-size-fits-all eye kit as far as I'm aware.

Quote - 1) I should find out how much merchants make selling eye textures before I decide what I'm giving away for free. Perhaps I don't realize the total market size? I've often wondered about that. I'm not going to change my mind if the total market for eye textures is only $500 a year. But if I could realistically expect to make $10K or more, maybe I ought to rethink this. Any opinions, people?

1. If I do give it away, I'm going to make a lot of enemies as I undercut the market, right? Opinions?

As far as Item 1, I'm not a merchant so I can't say how much you could make a year. But as for Item 2-- people can pose their own figure, but merchants keep creating pose packages, people can create their own light set up, again, there's still a market for lights. I don't see the market getting undercut from this.

Really the vendor you want to talk to is Face_Off (OcclusionMaster).

This is awesome. I have NO IDEA how to do things like these from scratch, but I can appreciate other people's work G

GREAT JOB, Bagginsbill!

I haven't looked at this Matmatic thing - is it Python in Poser or is it external?  If external, Java or Python would be the way to go for cross-compatability and GUI.  Is there a link?

The main Matmatic info page (the introduction)
www.runtimedna.com/mod/forum/messages.php

The download
loftydesigns.net/Beta/Beta/MatmaticBeta1.zip
Numerous Matmatic Links in this thread
www.runtimedna.com/mod/forum/messages.php

Attached Link: The node cult

Matmatic has not GUI. You write scripts - it uses the script to produce shaders - mt5 files and such.

The others gave good advice - follow the links.

So it's a Poser Python script set and it works on the 'matmatic' scripts (Python) to convert to shaders (as expected).  Basically, a shader programming library in Python to create shaders (as MT5/MC6 etc.).

You got it, Kuro. I made a library of objects, one for each Poser node. I also made a Color object and a few others. I defined syntax that would let you construct the nodes using Python constructor invocation. I defined overloaded math operators, so that, for example when you add two nodes, the result is a Math/Add node pointing at the other two nodes. 

Matmatic scripts are in fact Python scripts. However, the environment they run in is prepared by the Matmatic script first, so you can make a lot of extra assumptions about what happens before and after your script runs. For example, in pure Python, you'd have to do "from matmatic.nodes import *", among many others, but I've already done all those for you.

There is a bit more - the nodes do some simple local optimizations. For example, quite often, multiplication can be accomplished without a Multiply node, simply because every connection between nodes has a number or color on it that is a built-in multiply. Matmatic tries to find aggressive ways to use those. It will also do smart stuff, like if you ask for sqrt(x) and x is a node, then you also get a Math:Sqrt node, but if x is a number or color, it will do the sqrt in python and just write the resulting number into the shader as a constant.

I'm working on more optimizations. The most important is common sub-expression elimination. This will automatically reduce the node count when repeated sub-shaders are discovered. There are more features I have in mind for improving speed.

A financial model that would generate enough cash for you to run your life to the extent that you could liberate more of your time and talent to the cause of procedural materials in Poser ... we would all wish this both from the  basic satisfaction of seeing someone richly rewarded for his brains, persistence and common sense plus it would be better in the end for us...the product would be better. This is my pitch for you to look into making it a commercial product.

If you create a product that , for instance, produces eyes through shaders only, and is therefore model-independent, yes it might challenge the makers of eye 'textures.' It might force them to do you one better, and improve their own way of doing things to get a better result than you. Anyone who resents such a challenge...you should not worry about them, anyone who takes up that challenge....will in turn push you.  It's a good thing.

::::: Opera :::::

I will add....I have purchase 4 sets of face_off's shaders for skin, Apollo, Realskin, P6Realism and the mac version of RealSkin, plus Occlusion Master. That's about $75. 

I jumped at the chance to buy each one. I did not worry that this might 'undercut' or somewhat diminish the market for texture makers, since these shaders perform magic on default skin .jpgs.  I would say that since face_off raised the bar on the quality of skin, it has hopefully forced the texture/character vendors to do better. Some of them ship shaders with their characters.

::::: Opera ::::: 

bagginsbill, If you ever sell your works, I would buy them too! You rock!!! ^^ Shaders do render super fast! About color and Nodes -- I'm finding it tricky to control color in the nodes. I eyeball a color and plug it in, only to discover the color is too bright/too dark, too saturated, or the wrong hue. ^^;;; Is there a better way to control color? I have a chart I made with realistic eye, skin, and hair colors. Is there a way I can plug those colors into nodes for hair, eye, and skin? Thanks. bB

That's what I was getting at regarding some UI for it, possibly with sliders for each color item and other settings - Bastblack I know you are very experienced, so if it gives you problems, think about the average user.

might as well pile on ideas.....

as well as a UI which would generate shaders for Poser, if it could read shaders already out there in the market and 'transmute' them into Matmatic/PoserMaterialRoom language....wow that would be amazing.....

http://www.turbosquid.com/Search/Index.cfm?FuseAction=ProcessSearch&intStartRow=1&istSearchKey=&intMediaType=1&Submit.x=24&Submit.y=7

::::: Opera :::::

Let the feature creep begin!  Mwahahaha!

Backing up bagginsbill here a bit, I can't see how a GUI will be of much use for this type of thing.  Mainly, it's a language (in its own right) that allows you to turn basically (but not totally) mathematical concepts into Poser shaders.  The only GUI for something like this that I can pull from is LegOS and its clunky little blocks (which I quickly bypassed for real programming). :)

Now, there might be a GUI concept of use in this somewhere - it would be ultra neato keen (sorry) to actually see the results of what you are 'programming' and to have a little script editor there as well (say, script text on one side and a graphical box to show the Poser material result).  So you enter whatever code and then hit a button to display what would result - something like a 2D/3D plotting application (e.g.: Mathematica or MathCAD).  But that would depend upon the extent to which this can be achieved.

my preference would be for changes to the shader tree to have an effect on the actual actor affected in the viewport under full shaded display with software or hardware assist.

Shader effects show up in the preview. You change something, it refreshes (optionally automatic) and you see the difference

::::: Opera :::::