Planetside Software Forums

I would like to open a new Open Project: Voronoi

It has been said that Voronois could be used to create good rocks. I haven't yet.
But there are quite some amazing things to do.

So go for those, put a Babelfish into your ear and try to understand Voronis, so we all might have some insight ,-)

For a startoff I have attached a Clip, a Node-View and a Preview.

Volker Harun Bruhn

P.S.: One thing for sure, Voronoi are much slower to render than Perlin

http://babelfish.altavista.com/

a Babelfish is available here.  ;)

interesting project, I'll take a look but probably won't get very far.

Okay ... it is getting difficult.

The Voronoi is scaled antiproportional to the distance to the origin.

Sitting close to 0|20|0 gives you large voronois, i.e. 10.000|20|10.000 is giving small voronois.
So first of all we need a function that disables this.

Any ideas?

Off the top of my addled head, could you use a transform node to move the origin to your new desired location?
Just a maybe...
Richard

Hi Richard, this is a very good idea. And it would make it possible to adjust the appearance to your needs.

One drawback would be an animation, another one a full planetary landscape ,-)

Okay,

a workaround that needs finetuning:
The Constant scalar that is connected with the Voronois' scales gets multiplied with the position.
See attached Node-image.

It is not perfect, yet.

Volker

Well,

I give up. The Voroni function seems to be useless for now. It is impossible for me to get the default heightfield area filled with even sized voronoi.

Maybe some time later.

Here's a first go hope someone can use

This looks good! I like it.

Most likely i don't understand what you are after,but if it's a main
goal to get evenly sized cells one could generate an image based
terrain with a hexagonal grid,for instance,and go from there.
Just a thought of a mathematical stoopey :).

I do not know what I am after, either ... so I try to be lucky again.

Or to put it in other words, I would like to see what can be done with the voronois - after that I might know what I am after.

An image map does work for low scale images - one thing that I am sure with is that the result should fit planetary scale ,-)

Wow this is great it looks very muck like the Giants Causeway.http://www.risingroadtours.com/toursaints.htm (http://www.risingroadtours.com/toursaints.htm)
Richard

want this is file sharing section???

I have no finished file to share - it is just a very open project ,-)

Restarting with the Voronoi-Vector noises which look promising.

And it works very fine ... more to come later ...

Edit: Attached image

Here's a though for the use of the voronoi 3D cell scalar. It can be used to change the colour of every voronoi rock, like we have in the fake stones. Every rock a different colour. I haven't tried this yet and I should spend some time on it but I know it will work  ;)

Love that last image Volker. Very nice Voronoi.

I certainly plan on working with voronoi more and will post any results here.

My mind is boggling already. Every voronoi rock a different colour, well why not every voronoi rock a blend of different materials.  :o

And here it is:

(http://img522.imageshack.us/img522/6748/voronoipatchworkvn3.jpg)

The graph:

(http://img142.imageshack.us/img142/2306/voronoipatchworkgraphxj8.jpg)

101 other uses for this. Think about it. ;)

fake stones shader RIP. :D

Once again you are heading off in new and interesting directions, having read this thread I couldn't wait to experiment.

I just wanted to keep it simple and see what sort of shapes I could get, after initially just adding a mess of nodes all over the place I started again and created a simple ordered set (Clip attached), that would let me try the different combinations.

Volker that little group to create Pi is way cool  ;D

Here are some interesting shapes along with the node connections.  Nowhere nearly as advanced as where I guess you are but there may bo something in there  ;)

Volker I can't start to imagine how yopu got those oblong "cobbles".  :o


EDIT  Clip removed; fixed version further on   :-[

Wow, awesome work Mr_Lamppost.

These are so cool.

The first one is an effect that I have been wondering how to do for a while, I want it in one of my other projects  :)

That multi coloured cell idea looks interesting.

I mentioned elsewhere that I thought this should be possible, hadn't tried it, in connection with applying colour variation to a population.  I may be wrong but I suspect that the instances in a population are placed at the centres (Not sure if that is the correct term), of Voronoi cells.  As long as the instanced object is smaller than the cell it should be possible to use a Voronoi Cell shader with the same parameters (spacing, Seed etc.), as the population as a blend function allowing colour variation.  I have not tested this and am unlikely to find the time anytime soon especially if you keep distracting me with WOW things.  ;D


EDIT  removed broken clip   :-[

These are quite clever. I wonder if you take Heightfield erode v3 to the "vori_test_4" you could have a type of Dome Dune/Network Dune.
Also take the mutli-colors to "vori_test_1" and you could have soft rounded river pebbles in a river bed.

Quote from: Volker Harun on October 24, 2007, 04:54:04 PM
And it works very fine ... more to come later ...

Edit: Attached image

your work always fascinates me....

Third time lucky  This clip file has the missing node  ;D

Sethren:  The only way you will know is try it and see  ;D

Quote from: Mr_Lamppost on October 24, 2007, 11:08:03 PM
Sethren:  The only way you will know is try it and see  ;D

Agreed. I had to clean my PC out today with my air compressor with all the dust and ash in the air she is running hot and slow. Soon though.

Voronoi cell to merge two displacement shaders:

(http://img206.imageshack.us/img206/582/voronoidisplacemixls8.jpg)

Vector multipled:

(http://img452.imageshack.us/img452/6633/voronoivectormultipliedxf9.jpg)

A bit off track but crazy voronoi sine distortions of position and scale, Fractals could be used:

(http://img167.imageshack.us/img167/320/voronoisinesge2.jpg)

I have no idea on where to place the Heightfield erode v3 in the Voronoi group.    ???

Okay! As soon as I turn my back to have a nap, this thread goes wild - heh ;D

I love those results!

Mr.Lamppost, you might want to use a power scalar after your cos/sin functions ,-)

Efflux, ;D
and how do your cells work at larger distances to the origin?

Volker

Efflux,
You should use the Vector A for the displacements and Vector B for the colouring, just plug a Luminance to Scalar node afterwards.
Use a colour adjust after the Vector B!

This turns out to become wonderful, one prob is to combine different scales of Voronois - for me ,-)

Yeah, that's the problem.

Everything is cool here and working. The pi thing is really cool. Lots of shapes. Mojo can do this but we still need all this built into a fractal in TG2. I will experiment with colouring but test rendering a few things at the moment. The masking of individual stones is not really possible (or at least not for displacement) because of the shapes we now have with the pi function but not a problem. There are other ways to weed out some stones. If we ever get a proper voronoi fractal then all this tweaking of the basis must be included.

I do not use that Pi-Thing ,-)

Somewhere in the Image-section I found some useful stuff - you might be aquainted with ;D

The Nodes:

He he. This is exactly what I was thinking.  ;D

Some tests:

(http://img141.imageshack.us/img141/6426/voronoidistortionstj9.jpg)

You see here in the second image how it's screwed up with the ray traced shadows. They go wrong with lots of geometry extremes but this should not be so. The little black patches.

Perlin fractal distortions of the voronoi of course but we want a voronoi fractal doing it's own distortions.

They are nasty ... turning up AA helps a bit.

I'll go for the distortions, soon.

That new graph should be capable of some wild stuff.

My distortions are actually too much. I need to fine tune it a bit.

Distorted voronoi vector and cosine = Sand dunes. 8)

I bet that Matt, Jo and Oshyan and those who stay undercover do say in chorus: 'At last!'

They must have been very patient with us ;D ;) :)

Amazing stuff, guys. Respect!

You guys are really pushing the envelope on this, serious respect. Keep up the good work and PS bring on the macro facility so those of us with smaller brains (ie me :) ) can play too ;)
Richard

Okay - for the Voronoi B Vector, use a Length to Scalar Converter, you will not need a colour adjust afterwards and everything is smoother.

OK, I will try that.

Here's another old trick brought in. Already discussed on the forum. ;)

Also, distorted voronoi B vector and sine and those nasty ray traced problems.

We've only just begun. ;D

(http://img132.imageshack.us/img132/8247/voronoivectorsinedr2.jpg)

I would really like to understand those different Voronoi-Nodes:

• The Voronoi Diff Scalar, why does its amplitude get smaller with increasing distance to the origin?
• The Voronoi Cell Scalar, which cells get effected?
• The Voronoi A Vector is clear for me.
• The Voronoi B Vector seems to start at the Voronoi's random point and goes to the corner?!?

And why do the Scalars and Vectors not match?

Can this be utilized?

Sigh ...

I don't know the answer to those. The first point I have just learn to utilize the way it is, the second point - I'm not sure what you mean about that and the last point I have not noticed yet.

You can easily see that the B-Vector is building those ridges. Though it has some similarities with the Cell-scalar, it is quite different.

How do you utilize this scaling issue of the V-Scalars?

Nice images, by the way ,-)

Not using Pi, as I did recently, gives results, that are not understandable.
I hope vice versa is true ;D

I've stopped using pi now. Results are different but the same general effect. However it's now more the effect I would expect as I've seen in Mojo. I tried the length to scalar. It works to a certain degree but the colour adjust gives you simple editing power.

I often multiple by decimal places if I need to - which isn't always the case depending on what scale you're working at. It's just that multiply is more versatile. Sometimes you need to use both multiplying and dividing.

I'm still orientating my way around how things need to be calculated (scales etc) so I haven't formulated in my head yet how everything works.

Some of this stuff here could be ultized in Mojo to do things nobody has ever done in Mojo because Mojo does some of this stuff for you but you can still get at the right parts to do what we are doing here. Then I think of the Mojo UI and the idea goes out my head  :(

I'm seeing what you mean now about scalars and vectors not matching.

Also, I think the pi function is still useful.

 :o  <--  Every time this guy's eyes grow, I'm burning 1000 brain cells. 

Thanks.   ;D

Calico ... I am on a negative brain-cell counter ;D

Efflux ... it looks like the Dot and Cross Product of A and B vector might be something you like ...

Edit: Attached is a dirty quick render of the Dot-Product

That looks interesting except for the sharp angled displacement where the cells edges don't meet smoothly. We have to avoid any functions that cause that.

Agree - those steps are a hazard!
I think that highest priority would be to create a simple function too smooth these.

Edit: It is one of those moments that my heart starts pounding like meeting my wife the first time!

1. Make this Get Position stuff for an A vector
2. Make a Dot-product of the A Vector with itself
3. Smile!

Works the same with the B-Vector ;D

I was thinking about that but I don't know how to solve it. This problem crops up all the time (It happens when you blend terrains as well which is bad news). Functions which cause a straight up addition of the displacement and a sharp edge. They are no good.

Look at the edit above ;D

OK. I will try that. I'm working with luminance to scalar then to a colour adjust now. I'm not sure about this vector rebuilding though. OK for sine but I'm having a few problems with voronoi. Pi was smoother.

Attached is a view of my workspace with the previews.

Most interesting is the preview of the Dot-Product (the black and white image on the left side).
This is what I love ,-)

Here we go:
For the following picture I used the Dot Product of the 3D-A-Vector with itself (which produces a scalar). The Scale is ~300.
One displacement goes 200 positve, a transformed displacement goes 200 negative.

Again just a very rough and dirty quick render, that line in front is a luminous surface in a shadow (but the sky is nice) ,-)

Volker

Something that I noticed before, the Transform Shader does not seem to have any effect ... can anybody confirm this?

Multiplying the DotProduct with Pi and running a Cosine creates Blobs  8) ;D

I haven't had time to build alongside of you, so I can't right now.  I can always try a Transform Shader in what I'm working on if it helps.

Quote from: Volker Harun on October 25, 2007, 11:22:19 AM
Something that I noticed before, the Transform Shader does not seem to have any effect ... can anybody confirm this?

The Transform Shader does work, it just may not work in the way you need it to, or for what you're doing. ;)

- Oshyan

I have failed at getting the transform shader to work with functions, but I assumed that it was just me.  This is some pretty awesome stuff, you guys!  I'm wishing I wasn't so busy with classes so that I could help contribute ;).

Oshyan ... so you think that a 'Get Position in Texture' would be wiser?  :-\ :) ;)

I'm honestly not sure if the Transform Shader works with Functions, so that may be the key.

- Oshyan

Oshyan, oh it does and did ... The Position in Texture was the key ;-)

Ah yes - it needs to be provided the right info to function, yes?

- Oshyan

Ah, good to know!  Thanks for finding that.

I think a get position would work when a displacement is made befor the function. Just a guess as it is the first time that I used the Position in Texture node.
Thanks for pushing me this way - even if it was not on purpose.

OK, I'm back.

I never tried the dot product in this set up. I will do so but what I want is an output from dot product providing a usable value of angle between light source and surface. I could not get that working.

These ray traced shadow problems are bad. I've seen a lot of renders ruined with those. For example, I've even seen it a Luc Bianco image that was otherwise a great image. I will be testing that problem out further.

I use multiply vector in a lot of these function cases where I need to transform.

You mentioned this Dot-problem earlier ... this might be the next goal.
Here it works nice, it kills the steps and minimizes those shadows.

The multiply vector is of course something I ignored ,-)

Here is another render which highlights a few problems. I went reasonably high with the quality and used specularity because that shows up problems more. The Pi function got smoother results I'm sure of it (even although I have distorted loads in this one). This will be because of the way position coordinates are pushed around in the vector to scalar method which doesn't happen with Pi. We end up with steps again, even if minor. Pi is more limited though (or is it ?). Maybe Mojo uses Pi because all the results we are getting with Pi are built into Mojo and not the other things we can do with the new graph. I don't understand yet how you have utilized that dot product but I've yet to try it. You can also see those black patches or ray traced shadows but they do still appear even with minor geometry distortions. At small detail, the ray traced shadows seem to break to almost 100% failure. Well, there are often cases where in a less distorted landscape you'd still have your picture spoilt because of this. I'd expect problems at extremes but TG2 shows problems even with moderate distortion. Mojo got similar problems with steps in geometry and I've seen those black patches a few times but Mojo's underlying terrain displacement and blending is totally faultless and beautiful. You'll notice that in Mojo renders. The terrain flows beautifully smoothly even when masses of distorting and blending has gone on. Only surface displacements gave problems and in many cases actually way more bad artifacts than TG2. I'm always comparing to Mojo but Mojo is the only other comparable app and I know it so comparisons are inevitable.

I'm going to experiment with Pi again and better understand what it can do. I only know it's very basic function.

(http://img151.imageshack.us/img151/7160/voronoistepsproblemuu7.jpg)

You're right. Dot product gets smooooooth results, eliminating those angles. What I'm getting now are the best forms I've ever got from TG2 so far. ;D

Shame about those shadows though  :(

I think you are going to like what's coming next and we ain't finished yet.  ;D

Next image:

http://img528.imageshack.us/img528/2879/voronoi100uc8.jpg

Another one coming soon with a graph.

Next:

http://img90.imageshack.us/img90/1060/voronoi120vb7.jpg

Graph:

(http://img147.imageshack.us/img147/2007/voronoidotproductgraphhk2.jpg)

There is another layer utilizing two colour gradients but no need to show that. The gradient has already been discussed on a thread. Now you might see why I want gradients.

last time I tried two merge shaders....it gave me funny results....

how did you manage to pull this off...

Quote from: dhavalmistry on October 25, 2007, 11:43:52 PM
how did you manage to pull this off...

^--- I do not understand this

I like your renders, being short on time now, more to say later.

Attached is an overnight image:

Very cool. Looks a little like some stuff I was trying to put in this sine world that got abandoned - maybe you'll see something from that eventually. I was holding back from showing the WIPs until it was finished but it got too complex. I'm looking into how this these new voronoi rocks fit into bigger scenes and it's looking pretty good. I have one test landscape but one other total extreme shot that I'll probably post here.

doing something similar now, starts to make fun, hehe


volker, that's great, any chance that you share a view of the node network?  ;D

It's probably not much different from other things we've been doing in the thread but heavy distortion i.e. a function plugged into the voronoi to twist it around. It can be used before the input node or sent to scale like the power fractal going into the voronoi on my graph.

One more render. Is this still a voronoi thread?  :D

We can't see the voronoi cells anymore. I have some other tamer ones that are a bit nicer looking but this monster is worth seeing:

http://img90.imageshack.us/img90/9774/voronoi150pu9.jpg

@ Efflux: In the node network I do not get the point to split the Dot-product into the axis.
As far as I know (Nikita's twisted spire thread) TG2 converts a scalar to vector automatic if needed.

@ Child@Play: Go on for it, The Node Graph is attached.

There could be a point if we wanted to alter each scalar differently. That was what I was doing with the sines but it may be different now. I will explore that.

Yes, that was my basic idea for splitting it with the sines to enable different functions on each scalar but it's not necessary for what we've done here so far.

Another interesting point:
Take any Dot-Product, make a Complement Scalar and subtract the Dot-Product.

Nice  :)

yeah, really nice  :)

but render times go up for me when i do that

I'm changing that graph. I'm not sure the dot product is always desirable either but I'll leave it in that graph otherwise the picture will not be exactly as the graph is. It does have a nice effect.

OK, graph is fixed but I actually want to experiment with scalars. No need to make it look more complicated though.

The Dot Product is only nice when using a Clamp afterwards. You should check that.

About the tough shadows: There is a soft shadow option, that helps a bit, too ,-)

Soft shadow helps but it sends render times through the ceiling but OK for final render I guess.

I'm experimenting with different scalar functions now. I will probably post examples. I'm also building up secondary displacements a bit more. They are very small or too large in those previous tests. It was just a surface I already had.

This is mindblowing.

I'm tweaking about with scalars. That's cool but once you start building up secondary surfaces a bit more it just starts looking stunning.

I put myself in line with others and would like to have a sneak on the nodes ,-)

We're almost there now Volker. Awesome rocks. OK, not entirely realistically formed. A voronoi fractal would help with that but in another sense beautiful looking. We have the graphs for this basically sussed now. It's just a question of tweaking around to get all sorts of shapes. Bring in other layers etc.

One thing that really cheers me up is the ease of this functions and the ability, possibilities of shading :)

It's the detail. Looking at Mojo renders now after working on these rocks and giving them some more surface, the Mojo stuff looks terrible although the forms are good. In a bigger scale Mojo is very good. that's because of the fractal variety.

The ray traced shadows are OK once you build up secondary displacement. I'm just tweaking around here trying to make it look good with shadows then I'll post.

I am rendering - the fourth time - now I want to be patient ,-)

Soft shadows- 4°, 9 Samples - takes about 1.5 times longer.

Is there any way to lighten down the darkness of shadows. I know GI does that, at least some times but in this case I've got very dark areas. I turned the sun up but still it's very dark gloomy shadows.

Anyhow, I think I'll go for a reasonable quality render with this next test.

Very, very nice results so far! Respect for your patience and your really good ideas!
I see, there is a vast variety of applications we can use voronis for.
I really like this topic and watch it since it has been started, because I also did a lot of different voroni tests within the last months.
Just wish I had more time to experiment (could bite my ass for that).

Last time I tried a scene with really big voroni cells and displacements for overhangs or granite massifs.(used values up to 800 for scale and 500 for
the displacement shader)
I have no test-render heres at work, but results looked similar to this (the scale in my scene was bigger but the rock-forms looked like this):

Ironshirt! I am looking forward to see your results!

Efflux: Add a Cirrus cloud layer, increase depth to ~20-40, increase coverage to 2. This should darken the whole scene. Increasing the exposure will light up any part of your scene.

It appears that TG2 simply can't render ray traced shadows properly from the secondary surface displacements at all. This is the problem. If you have no displacements you get no problem. Take a displacement that is not hooked through a compute terrain then hook it through, you'll see how the shadows from the displacement not hooked through are useless.

If the ray traced shadows can't render these displacements then they shouldn't even be trying to but just render from the terrain.

Hopefully the renderer that is built right now by Matt will get rid of these.

Lets hope because even if we just had ray traced shadows coming from the terrain and ignoring detailed surface displacement shadowing it would actually be better.

I do not like to bet, but I guess that Matt would recommend to use higher GI-settings ;D

I think the problem might go away when you do huge renders at higher settings. On that "Functions" render I did, I had these same problems yet I don't see them on the 3000 x 1500 final render. I remember them appearing in the displacement cracks of test renders.

Well, at my new work, there will be a 12xQuadcore Server idle over night. Maybe I can crack in? ;D

It's hard to work out what is actually happening. The fact that bigger renders of the same scene show less of the problem suggests some kind of level of geometry squeezing causes it because it's definitely much worse if you distort things.

I've tried this next rock test and the shadows are dire so now I'm rendering it at twice the resolution to see what happens.

I dumped everything done so far :)

Starting back with one of my Perlin-Spires and recreate the surfaces using voronoi and perlin ...

May take some time, as I want to go final ;D

I'm waiting for a voronoi stones render. A waste of time really because of these shadows issues but I'll use these ideas. I'm working on atmosphere a bit to try to create more distance and space. The default atmosphere settings are just about useless for anything.

For example, if you look at that latest Armand's picture it has huge depth and space. Quite a few Mojo pictures are like this but it has nothing to do with the app. TG2's atmo is better. It's to do with the huge planetary vistas so some Mojo users have simply learnt to set the atmosphere better for this feeling of space. I also think that any graininess in the atmosphere is out of the question for this spacious effect.

The main problem with TG2's atmosphere is that it does not count the scale you are working at in ,-)

As soon as you work on sea level with reasonable heights, all you need is to adjust the atmosphere's density and height. Of course some lighting conditions need more work. But as long as the scale is fine, it works for me.

Yes I'm finding the scale is important. Atmosphere goes wrong if you go out of scale. One scene I'm working on now looks pretty realistic so I think I'm getting somewhere with the atmo. It's a clear atmo which always makes nicely surfaced near objects look good but it can be tricky to tweak those atmos for a good distance look.

Here are my 2 cents....


a realistic approach using voronis......for some reason I cant get Volker's surfacing technique get to work....I must have forgotten it.....

anyway...ignore the surfacing.....

The surfacing is fine.  The green is an offish color (depending on the context)...maybe.  Anyway, I like what you've done here.

Quote from: dhavalmistry on October 26, 2007, 12:21:35 PM
Here are my 2 cents....


a realistic approach using voronis......for some reason I cant get Volker's surfacing technique get to work....I must have forgotten it.....

anyway...ignore the surfacing.....

great dense jungle!
Kevin.

Mistry, your voronoi works good here. As wished, I ignore the surface ;D :) ;D

more......I call it Grey Matters!!!

dhavalmistry, these ones are much better. You see the rock forms without so much surface obscuring them. You should try sending the output from the voronoi to X,Y and Z with to scalar nodes. You build it back with a build vector at the end but it means you can process X,Y,Z differently. I sent X to sin scalar then power scalar and Z to cos scalar and square root for this next image. Just throw in functions and see what happens. This image would not render with shadows. I had to abandon that. It's a big problem shadows in TG2 are useless with small surface displacements. This totally diminishes the feasible use of displacements entirely.

http://img142.imageshack.us/img142/3640/voronoi160oc1.jpg

Attached is last night's render, I am not yet satisfied with. More to come next week

The towers look great. Surfaces are great as well. I'm not sure about the atmosphere though but like I've already said I think some serious tweaking is needed for TG's atmosphere - well away from the defaults. You'll see my results soon in a new planet. I'd also watch you don't get too surreal but technically it is a great angle to learn about TG's limits. I'm actually coming back from that a bit now. You'll see what I mean. Fairly surreal in terms of rock forms etc but texturally and atmospherically real looking. I think this works well, convincing the viewer that it's real. I've taken close ups from this planet's rocks and you'd swear they were photos even although the rocks are absurd shapes.

Also. I don't get this shadows thing. It's going to take some time to fine tune that. My planet now has much fewer problems than those rocks before but still has the same rough texture.

See you.

AyeAye - changed the atmosphere (which was the most tweaked atmo I ever had) and went for PoV and surfaces ...:
http://www.renderosity.com/mod/gallery/index.php?image_id=1549919&member

for some reason..it reminds me of terminator....

It's a boneyard for prehistoric fossils.   ;D

Good job, Volker.

Quote from: Volker Harun on October 25, 2007, 10:36:11 AM
Attached is a view of my workspace with the previews.

Most interesting is the preview of the Dot-Product (the black and white image on the left side).
This is what I love ,-)

what exactly does "power scaler" do here??....I am having little hard time to understand

The Power Scalar gives the power of a number by the scalar.
If the scalar is 2: 0.1²  0.2² ...
If the scalar is 3: 0.1³  0.2³ ...

The result is that values near zero become very small and values near 1 get only a bit smaller.
This widens the black area and makes it a bit smooth.

Great job Volker. That's what I was kind of thinking of. If the atmosphere gets to wild it doesn't looks so good. This version is really cool. Much better. Your surfaces are fantastic.

I like them both,but the first one is the better as far as i am concerned.
That goes for the terrain and the atmosphere,which had a better mood
in my opinion.

The second one has more of a feeling of space and distance in my opinion. I'm guessing that it has a lower exponential for the haze. I've been experimenting with this. The default haze in TG2 makes you whole atmosphere very thick and hazy but by dramatically reducing it's exponential height you can get a clearer atmosphere and see more of your scene and use the haze to actually get a very low mist effect instead of having to resort to 3D clouds. Some of the Mojo users were saying that TG2 renders don't have a feeling of space and grand scale. One reason is simply getting used to creating scenes on a large planetary scale, the other is tweaking the haze to open up the atmosphere. There are some other reasons but these are all not so relevant.

This gripe about the haze in the TG default atmosphere has been going on since virtually day one.

There are loads of ways to tweak it but here's a clip (Single node) ;D  for the atmo. I use as default. 

Concerning the very first post. I made a logical mistake (or to put it in other words: I can't be lucky every day  ;))

The very first powerfractal that is multiplying the Get Position has of course a larger effect on large values ... The PF should be multiplied with a vector. This result is to be added to the Get position  :D

Now it works independent from position  :)

(P.S.: And thanks Dude for showing me the way!)

Volker

Quote from: efflux on October 26, 2007, 07:34:45 AM
Lets hope because even if we just had ray traced shadows coming from the terrain and ignoring detailed surface displacement shadowing it would actually be better.

It's not quite as simple as that. If the additional displacements cause the surface to go below the "terrain", they would be plunged into shadow where they shouldn't be. I really think you would like that even less ;-) This could probably be avoided by calculating the amount of shadow reaching a point which is different from the actual displaced point, but then shadows from other objects would be wrong. The only way to do this properly is to account for all displacement. The new changes I am working on should provide 100% matching between what is visible and what casts shadows. I can't give you an ETA for that yet.

Matt

The default atmosphere gives a fairly good "Earth-like" result at normal scales. Obviously if you go outside of those norms or want more "extreme" results you have to tweak. I think people actually often want an *exaggerated* impression of distance and scale. TG2's atmosphere is very powerful and flexible, it's just a matter of tuning for what you want.

- Oshyan

Quote from: Oshyan on October 31, 2007, 06:05:32 PM
The default atmosphere gives a fairly good "Earth-like" result at normal scales. Obviously if you go outside of those norms or what more "extreme" results you have to tweak. I think people actually often want an *exaggerated* impression of distance and scale. TG2's atmosphere is very powerful and flexible, it's just a matter of tuning for what you want.

- Oshyan

I'll have to agree with Oshyan on that one. I think it may also be because people don't often place the camera at a realistic eye level because they want to see a lot of the terrain... but expect the same clarity in the foreground as if the camera was positioned at eye level.   

Thanks for the replies here. For some reason I do not always get notification about replies on threads so I missed these.

I agree that TG2's default atmosphere gives generally natural results, at least definitely like the kind of atmosphere we get in the UK - thick and misty. I don't know about other places though. I have experimented a lot with this and found that the open space kind of feeling you see in say Mojo renders is because there is less atmosphere or haze going on at least certainly at higher altitude. If you want this it's quite easy to just reduce the haze exponential hugely and increase it's density. This opens up the space immensely and the low haze gives a good sense of distance but you also need to set higher samples so as not to get grain if you're creating thick low level haze, like you often see in Mojo.

I will wait to see how things go with the shadows.

Thanks.

I agree that the default atmosphere is a good approximation; at least on some days.  I prefer to start with a clearer version (As posted), and then thicken things up again as required.

If I am looking for a natural look to a scene I try to place the camera 1.7m above the ground to give a natural eye line.  1.7m is natural for me, I am quite tall  ;)

Yes, you're right. That's another issue. You need to bear in mind normal scales i.e. camera placement otherwise everything else is meaningless. You have to be careful with that when developing anything. A rock that is supposed to look 1m should be 1m. It's easy to get all these things out of scale.