This is a 100% procedural cloud shader for my environment sphere, not a photo.

There are only two props here - my environment sphere and the Poser Ground plane.

It is a new all-time high for node count. It uses 744 nodes.

I implemented a volumetric ray-tracer entirely with nodes. There are two layers of clouds here. One at 15 to 20 thousand feet, and the other at 25 to 33 thousand feet. The atmospheric haze is such that it blocks half the view of the distant clouds at 60 miles away. These are distances you can't deal with using props. But there are no props at such distances here. The environment sphere radius is just 750 feet. The appearance of distant objects is all because the perspective simulated by the ray-tracer using the nodes, which seem to be able to handle huge numbers just fine, unlike the rest of Poser.

The water shader is not very good because the bump is just one node. I need to work out the math for an infinite series of distorted trochoids.

Anyway, I thought you'd like to see it.

Be sure to click for full size.

Wow! That looks amazing!!!  But it doesn't hold a candle to you candle shader which is still my most favoured one that you have made. Well, that and the glass globes :)

THAT IS INCREDIBLE!!!!  SIMPLY AMAZING!  Definitely a must have for any serious shader fan. 
But, wait a minute, isn't that Clover swimming in the distance?

 lovely

744 nodes ?  How do YOU keep track of them all....

Phew !

Water shader is going to be interesting as well.

The grand scale for Poser scenes, is finally at hand !

I think it looks quite good. If I were going to be picky I would point out a couple of things. Too much yellow in the darker parts of the clouds. Perspective looks wrong near horizon, where clouds appear too small. Would also like to see an overall bigger cloud size.

I'm seeing a pattern in the water and I agree with the clouds looking too small on the horizon line.
Not bitching just to bitch but that's what sticks out to me. Nice work.

Here is with a 100mm lens.

Unfortunately, this reveals why I used the wide angle. I'm getting visible artifacts from my sampling of the clouds miles and miles away. If I reduce the sampling to fix that, I'll need a couple thousand nodes.

As for the yellow - there isn't any. The bottoms of the clouds are neutral light gray. The tops are a little bit blue, which would make the bottoms appear yellow. I thought the tops being lit a bit from the sky should be slightly bluer. Is that wrong?

Anyway - when I release this, the colors will be adjustable, as will cloud size, density, puffiness, and a bunch of other things.

BB, will you ever out-do yourself?

This is incredible!

No one is going to mistake the clouds for a photograph, but not bad. Good work.

Coldrake

amazing, and an amazing volume of nodes too....

700 nodes? rendertime is very long  right?

you did this with matmatic right?

it will be interesting when you explain how you made them.

'' I implemented a volumetric ray-tracer entirely with nodes.''*
holly s..... thats insane

That is rather incredible.
Randomizing the cloud size more would greatly improve the realism I think, although one of the biggest problems with an overall procedural technique like this is really being able to predict what will appear and where. What I mean is, when I'm taking a photo or composing a painting I look for specific shapes that will lead the eye around the image. In this case I might be looking for cloud formations that nicely "frame" key features in the fore or mid ground, or which particularly add a sense of depth by leading the eye into the distance.
This is difficult with procedurals, unless you can define specific areas in which the effect will take place.
In Vue (when I had it running in V5) I could place specific cloud objects (cuboids whose final rendered formations would be procedural) where I needed them, and had more random and lighter higher level cloud formations as a general backdrop to these. Might be a direction worth considering.

I'm very impressed though - yes, render time(?)

 BB,

Every time I read one of your posts I am astonished!

I would like to have .01% of your ability to come up with stuff.

:::Bookmarked:::

BTW, Thanks for sharing all your talent and abilities.

Quote - No one is going to mistake the clouds for a photograph, but not bad. Good work.

Coldrake

Oops. Silly me. I was very punchy when I wrote that from lack of sleep. (Not clouds work - real work kept me up late nights lately.)

Of course it's nowhere near photorealistic. I meant to say:

This is not a photo-based solution. One technique for cloud rendering is to take a photo of a cloud puff and use it to drive an algorithm that generates thousands of slightly morphed copies of it.

Quote - 700 nodes? rendertime is very long  right?

you did this with matmatic right?

Yes, of course. Even if a human could keep track of all that, the material room becomes so slow with so many nodes that each change would make you wait several seconds.

I generate the nodes with matmatic. There are so many because Poser shaders do not have looping procedures with variables. So, matmatic does the looping, and for each iteration it generates a set of nodes. In this render, there were 30 iterations for each of the two layers, so 60 sampling layers. Each sampling layer requires 12 nodes - that's 720. The other 24 nodes are for some control parameters, color choosing, and a few basic coordinate transformations that feed into every iteration. So the shader really is only 36 nodes, but part of it is repeated so many times to get the volumetric sampling.

Render time is longer than a simple photo, but not nearly as long as ray-traced shadows on transmapped hair. These images rendered in about two minutes.

Quote - That is rather incredible.
Randomizing the cloud size more would greatly improve the realism I think, although one of the biggest problems with an overall procedural technique like this is really being able to predict what will appear and where. What I mean is, when I'm taking a photo or composing a painting I look for specific shapes that will lead the eye around the image. In this case I might be looking for cloud formations that nicely "frame" key features in the fore or mid ground, or which particularly add a sense of depth by leading the eye into the distance.
This is difficult with procedurals, unless you can define specific areas in which the effect will take place.
In Vue (when I had it running in V5) I could place specific cloud objects (cuboids whose final rendered formations would be procedural) where I needed them, and had more random and lighter higher level cloud formations as a general backdrop to these. Might be a direction worth considering.

I'm very impressed though - yes, render time(?)

Gotcha. I've seen Vue's clouds - beautiful stuff - what I aspire to do although I doubt I can get it that good.

I'm still working out the principles of how this works. The interesting about it is that the sense of depth does not come from geometry, although the placement does. Meaning, if instead of a full 360 environment sphere, you used the shader on a simple rectangle, you could position it anywhere you want and that's where you'd see the clouds. By changing the algorithm to confine it to just one puff or some other desired shape, the individual rectangle can be dialed to produce something specific, and by positioning the rectangle in the scene, to put that in a specific place.

Note that positioning the prop would only control the viewing angle where the puff would appear. Its actual 3d look and it's apparent distance from the viewer would be from the shader. You could put the prop just 20 feet away from the camera, but the cloud would still appear to be a few miles away. Think of it as a little window into another world where there is a cloud perfectly framed by the window.

Can you create more dramatic lighting effects with this kind of setup (rim-lit clouds)?
In any case, I'm mightily impressed with the render time of just a few minutes with all those nodes.

In your most recent posted image ( with the artifacts problem)
you no longer can  see the  3 or 4  tiny dark bits  right at the point where the water meets the sky.

In the earlier posted images, i saw these and wondered if those tiny dark bits were supposed to be visible, i wondered what you were going for ...but i like the fact that they are missing in the most recent image

Quote - In your most recent posted image ( with the artifacts problem)
you no longer can  see the  3 or 4  tiny dark bits  right at the point where the water meets the sky.

In the earlier posted images, i saw these and wondered if those tiny dark bits were supposed to be visible, i wondered what you were going for ...but i like the fact that they are missing in the most recent image

Those dark bits are Poser reflection rendering artifacts. They are showing up on the ground plane where it touches the environment sphere. They are 750 feet from the camera and Poser has numerical accuracy problems with this combination of extreme overall distance, extreme close distance between reflected and reflecting surface, shallow camera angle, and the bump pattern distorting the normals to create the waves.

I think they are still there, but in the zoomed in view, they are outside our field of view.

I'll sort out the tweaks necessary to get rid of those before publishing. That's why this is a WIP!

Quote - Can you create more dramatic lighting effects with this kind of setup (rim-lit clouds)?
In any case, I'm mightily impressed with the render time of just a few minutes with all those nodes.

I think so but it will be tricky as all getout. Right now I'm just trying to get the overall effect to work. I'm not happy with the cloud function itself. I did this with FractalSum because it is the fastest of them all, but it doesn't produce the kind of pattern I want. The cloud boundaries are very diffuse here, even though the FS node is set up to do 24 octaves of noise!

I'm experimenting with fBm. This node is 10 times slower, and cannot be used in 60 layers without serious performance problems, but it has a much nicer detailed pattern to it. But, to use it, I have to solve the sampling artifact problem. I need to get the number of sampling layers down from 30 to around 6 in order to use that node. Unfortunately, the more detailed the cloud is, the more sampling layers you need. With the fBm node in the same setup as I have here, I get much worse sampling artifacts even with 30 layers. But the pattern is much nicer.

Ultimately, I want the shader to pay attention to the main light position. You should be able to move your infinite light, and this will influence 3 things:

1) The clouds will be lit from the right direction. At the moment, they are always lit from directly above.

2) If you are looking at the "sun", there will be a hazy glow around the sun. If there clouds between you and the sun, or even nearby, they will light up in full HDR.

1. If you move the sun to near the horizon, the sky color will change to a full sunset, and the clouds will turn gold or pink or whatever, just as in real life.

whait a minute. those clouds are lit with lights?

Yep.

what?????????????????????

so this is not just an effect. you actually build 3d clouds that are lit by lights?

this just got 200 times more awesome.

Well, those words don't quite capture it.

I build a 3D density field. Within that field, there is a lighting gradient. Remember there are no surfaces here at all. Everything is about scattering light within a medium of varying density. Light can reach into the cloud only so far, because it is scattered. Light arriving (directly) on one side of the cloud is stronger than on the other side. However, some of the light that arrives in the middle gets scattered to the edges, so the shadow side is not at all as dark as it would be if there was a real surface with opacity.

Also, when light reaches into the middle part of a cloud, but there is more cloud between you and that part, you have to decrease how much of that you actually see, because the light cannot scatter all the way back out.

I have not fully worked out all the math for this, nor am I happy with the density function itself at the moment, but I'm making progress.

Look really realistic !!!

I'm thinking ... If you've created two layers for the clouds, may be you can create two layer for ... a skin for exemple ! I say that but i say nothing ...

*Well, those words don't quite capture it.

I build a 3D density field. Within that field, there is a lighting gradient. Remember there are no surfaces here at all. Everything is about scattering light within a medium of varying density. Light can reach into the cloud only so far, because it is scattered. 

In other words you  recreated nature inside of Poser

as water vapor becomes more dense in  the sky, clouds are visible

nice

Because this density pattern is more detailed, my sampling rate is too low. I have to find a way around that problem.

But I like these cloud shapes a lot better.

this is more realistic.

why is realism always slower? can god not give us a break? 

But ... to get this kind of detail, I had to increase the number of samples. Each point in the render required samples equivalent to a ray-tracing depth of 100. Imagine trying to get Poser to do that with the renderer, instead of the shader, going through 100 layers of transparency. Impossible.

I was able to make each ray with slightly fewer nodes - just 9 instead of 12. So this is 922 nodes. This reaches my limit of tolerance for load times - it took 25 seconds just to load the shader.

Render time was not too bad - about 4 minutes. But I have some ideas about how to change my sampling strategy, which will let me increase speed by a factor of 10. If I can get this under a minute, I'll be happy.

this looks even better.
a little to complicated to render in poser since the renderer is not fast enough( i guess).
but still very amazing what you did.

Amazing work yet again, but wouldn't expect less. Could this be used to generate smoke coming from a lower object too?

Quote - Amazing work yet again, but wouldn't expect less. Could this be used to generate smoke coming from a lower object too?

I think it could. I'll have to think about that. I made a morphing smoke plume prop and shaders a couple years ago and posted it at RDNA, but the link to the zip file is gone. So I've lost that file.

Boo hoo.

Anyway, it wasn't as good as this is looking, so I'll think about how to do it better, the way I'm doing here with a ray-tracer built with nodes. The one I built before involved thousands of procedurally trans-mapped polygons and it sucked for rendering because Poser gets very upset trying to do dozens of layers of transparency.

Here is the smoke plume thread, in case you want to see it. But the link to the file is gone.

http://www.runtimedna.com/forum/showthread.php?t=31335

Absolutely amazing stuff, BB.

I remember reading that thread about the smoke at RDNA , and was wondering if that's how this started. When you did the fog on the ground I tried to make that into smoke with no success.

I have the smokeplume.zip if you want it bb. It's approx 1.4mb.

Thanks raven (and GeneralNutt, too, who also offered in PM).

I found it. I woke up this morning and I remembered where I posted it - on divshare.

This was before I made my own Google site. Now I don't have this problem anymore.

 Very impressive. 

Wow, BB, that looks like the sky outside my window right now! That looks realistic to me!:laugh:

Any thoughts to a cloud plane?  We have a ground plane, why not the other?  Would any savings of time be from a smaller cloud plane with the slower  fbm node on it?  Does the area of the env  sphere (in sq feet or poser units) affect render time?

Does the angle of the sphere of plane affect the view of the clouds, or is the effect independant of the angle of incidence?

This quite exciting stuff to my mind.

Is a procedural night sky with clouds, stars and a moon on the horizon?  Or Am I shooting for the moon with this question? :laugh:  If someone says you've got your head in the clouds, just nod and walk away smiling. :lol:

Using a plane for geometry instead of a sphere is no good because it doesn't reach all the way down to the horizon, even if its size is infinite. Using a sphere gets full coverage down to the horizon.

Mathematically, I'm actually implementing the clouds as a series of concentric spheres, not planes, because I'm applying the curvature of the earth to the simulation (yes, I really am). In my first post I was using an infinite plane, and that was producing bad results, because we were able to see clouds more than a thousand miles away. Depending on your altitude when you render with my latest setup, the farthest cloud you see is between 8 miles and 60 miles. Of course, at 60 miles, it's all haze, but they are there where they should be and at the right size because of perspective.

Moving the camera does affect the perspective a bit, but you are free to move the camera without too much problem up to about 400 feet from the center of the sphere. That's plenty of room for animation.

I already have a procedural star shader, also galaxies, nebulae, etc. The shaders can be combined quite easily. A better technique would be to set up two concentric environment spheres. The outer sphere would have clear sky, which could be day or night. The inner sphere would have the simulation. This would run a little slower, because there is a transmapping involved on the cloud layer, but it's only one layer as far as the renderer is concerned, even if it is simulating 100 layers. By layering the sky (or stars) and clouds in two spheres, you could put a moon prop between them. You could put a light source in the moon prop and the clouds will haze and light up near the moon.

You could also use multiple cloud spheres, for multiple layers of clouds (cirrus, stratus, etc). These layered spheres do not have to be at those actual distances. They just have to be nested by about 50 feet or more.

The possibilities are endless.

B for Bagginsbill.

These movable clouds render really fast, because they don't need to go for miles, so I just used 10 samples, and they are always artifact free no matter where you put them.

When are we going to be able to play with this?

Hopefully next week. I have a lot of real work to do this week and my daughter's high school garduation is this weekend, so i'm kinda busy.

I will post it when i figure out how to do fewer samples without artifacts.

Quote - Hopefully next week.

Looking forward to it.

Quote - Render with ray-traced blur shadows. Sloooooow as with any transmap in this situation. Took about 15 minutes.

now imagine that you could set a ''cloud'' light with DM shadows. it would give you some soft shadows  and the render time would be faster.

we need this SM he he  :) 

If we could meta-blob stuff like in Vue, these cloud props would really fly!  Really impressive stuff,BB!

Out of curiosity, do you have a SUN prop?  We can't see lights in Poser, but there must be some convincing work around.  I'm sure of it.

Nesting the spheres is a clever idea, that would light up the moon.  That sounds exciting.  I can't wait to get my hands on it.  I got tons of ideas to try out on it.

Quote - If we could meta-blob stuff like in Vue, these cloud props would really fly!  Really impressive stuff,BB!

Out of curiosity, do you have a SUN prop?  We can't see lights in Poser, but there must be some convincing work around.  I'm sure of it.

Nesting the spheres is a clever idea, that would light up the moon.  That sounds exciting.  I can't wait to get my hands on it.  I got tons of ideas to try out on it.

you want to render out the sun?

import a texture of the sun and add it on a prop. or on the ENV_sphere.