This is a way I've used to create stepped terrains. Not sure if it's the best way but it's relatively easy. There are ways to control displacement by blue nodes only but totally arranging the steps requires control of both the fractal output values and the displacement results used from this. Each displacement node here is the amount of extra displacement that the step creates. The constant scalars on the right are the ranges of the fractal output used for each step. The output being used from the fractal is colour. The displacement settings in the fractal node do nothing. Fractals do not output -1 to 1, generally. Often they actually output beyond these values. That all depends on various tweakings in the fractal node. Each terrain will be different. The file at present will chop off anything that goes beyond 1 (not anymore - that is fixed) creating flat topped hills. Not a major issue is many circumstances but I will look into fixing that. The colour is however unclamped.
The terrain is very smooth. You can increase the roughness or the contrast in the colour tab of the fractal but other functions can be applied to create detail. The more you create steep sided hills, the less the original fractal can create nice details on those slopes. Beyond a certain roughess and you will get masses of spikes.
The last red node is called Test colour. I put this there so I could hook the ouput from each smooth step node into that as a blend to create bright colour. It won't match the steps correctly but gives a rough idea about the whereabouts of the step you are tweaking.
EDIT:
I won't remove the above text because that's part of this thread but I've removed the file first put here. A better solution has been added below so the blurb above does not totally relate to the new file. The older file was too inefficient. Some info to do with this new version is on reply 44 in the thread.
I just re-uploaded the tgd. I realised that there was a subtle difference between the file and the render. The file wasn't as ideal a starting place. It's still not. It has a fault. Another file will be coming.
OK guys. This has become a bit convoluted now. I've uploaded another file again to replace the first one. The clamp pulls the flatness out from the last smooth step to displace the top platform by using values above 1 but that has to have the subtract so as to not effect the rest of the terrain.
Now there is displacement node that is different at the start. That really needs to be there. You can use that to push the entire terrain up if you want. It also provides a way to hook another fractal in for more detail.
Maybe I'll find another fault but I think this file works OK.
Just thinking out loud here, but would it not be easier to use a PF outputting scaler values, which you can use a modulo to split into levels of grey, that you then plug into a displacement shader? With a little enginuity, I'm sure you can smooth out the tail end of each verticle step, to kill hard edges.
Yeah, the smooth steps are in there because I wanted it smooth. This is really crucial because TG2 does't like sharp edges and anyway they rarely exist in nature. Your method may work. It all depends on exactly the results you want and how far you need to take it. You don't want flat steps and you need to control the displacement level of each step.
Here's a render from that file. I just roughened the fractal up a little and added some colour texture.
Awesome! Loving it! ;D
Hetzen.
I had a play with the modulo node. The problem is that we can't have any straight transition. I'm not sure how any nodes that do that kind of thing can be incorporated.
Ideally this is what we want. We want to control the input to the function which is the fractal's output so that we can choose ranges from that. Ideally these ranges need maths functions to smooth them. A smooth step style shape is ultimately the most useful whether created by smooth step node or by another maths function. The reason this shape is best is because it doesn't just create a smooth step within it's range but a smooth step to the next range. We could use another maths function which allows you to change this curve. That would be cool. Bear in mind though that functions which move away from smooth step create a visible join between regions but sometimes this is desirable i.e a sudden angled outcropping from a flatter step. Then we need to control the output of the function in terms of something that makes sense i.e. metres. My file doesn't make total sense in that the displacement is cumulative and then the last clamp confuses it further because that doesn't accumulate like the others but at least it kind of makes sense.
I'm not starting anything new with this at the moment but if anyone else wants to try different methods then that would be cool. Mine works if not totally elegantly. It's a case of getting things out there that simply work. I'll be taking a break and I'm just petering out from using TG2 now until I get more enthusiasm in the future.
Hi Efflux.
I've been reading your posts with great interest and learnt a few techniques along the way. So it's a shame that you're losing interest.
Quote from: efflux on March 13, 2012, 04:21:30 PM
I had a play with the modulo node. The problem is that we can't have any straight transition. I'm not sure how any nodes that do that kind of thing can be incorporated.
I'm not sure what you mean by transition. There are issues with the modulo approach, mainly that you have to build your world in scaler, rather than displacement (although that's being looked at). What would be good, would be to step modulate the layer spacing to get variation between step hights. I'm not sure how to do that atm.
EDIT - Forgot to mention
The below is looking at a scaler PF, which is plugged into a modulo, using fractions of 1 (which you could convert to meters). You subtract that from the PF to get your step. You then take that modulo value and convert that to a 0 to 1 ramp that is smooth stepped, and this output is used to mix between the PF and the step value, this then gives you your smooth edges. You can also look at the tail end of your ramp, to create distance between the steps.
How do you "convert it to a ramp"?
Hi Mhaze. Simple answer, divide the modulo output by its step value (input 2). So for example, I'm using a step value of 0.1, 0.1 divded by 0.1 = 1, the top end of your ramp. I was getting some nice effects last night with mixing between two sets of this function, with a PF modulating the modulo step value.
I don't have a release version of TG to be able to upload a clip. And besides, this is Efflux's technique, which I don't want to hijack. I just wanted to engage in a discussion on how to achieve this effect outside of using the strata shader, which Efflux has demonstrated with the added bonus of being able to have different spacings between steps. I may start another thread with a function set that has more options at a later time.
Cheers
Jon
Thanks for the tip about the Modulo Hetzen.
Yeah, I was wondering about the ramp thing but I sussed it out last night after a bit of time.
I don't think this is good for natural type landscapes unless we can modulate it but reading you last reply I guess you're already trying this. I also notice there is now documentation at Planetside for the modulo. I knew what modulo was but didn't realise how you could use to create repeats. Just goes to show how poor documentation doesn't help. It's way cool if you use it for stuff like this as long as you smooth step it because lots of sharp steps isn't too good in TG2:
Hey that looks great! Almost like lego bricks.
Yeah I made a point of looking at each blue nodes description a while back, just to see how I could use each one. Some of the vector stuff I've had to learn via google searches. That ontop of seeing how others like yourself apply these sorts of functions. There are a few I can see no use for, but that's more down to my ignorance than anything else. Dot product is something beyond me, as I don't know why I would use it.
As sad as it sounds, I like playing with blue networks. I guess it's my version of Suduko.
The modulo works well when you apply it to geometric stuff. That last image is just two fractals stretched out in different directions. I love the way the modulo creates nice rounded type blocks where it crosses over. If that was sharp it would look nasty. I've noticed a lot of stuff in the forums where people have used functions that ended up with lots of jagged edges. This never works well.
Actually I correct myself. That last image is one fractal stretched on two separate transformers. I'm on Linux so not looking at the TG2 file.
Dot product can be used to get values between vector angles to effect certain surface directions differently but I'm no maths expert by any stretch of the imagination. I just know how it could be used if you knew how. I used to use it in Mojoworld to make the sun direction effect colours differently. Other than getting slope values, dot product effects would be quite weird. Strange colour shaders that move from one colour to another depending on whatever vector values you are using.
Yeah, that's the part I don't get yet.
I know what you mean about hard edges, my squared rock experiments look worse than they really are. Infact, there are all sorts of curves on the extrusions to avoid nasty edges, as well as trying to avoid verticle sides (stretched textures). The problem arises when you apply it to 'natural' terrain. Still ongoing that one. But I don't mind hard edges at the base of features.
Yeah, hard edges are actually OK in places. If they are all over on rough surfaces then it tends to get nasty. I think since hard edges rarely occur in nature they tend to look nasty anyway. Then there is the problem of applying displacements over hard corners. It just accumulates into a terrible mess and TG2 has an excellent feature as well in it's smoothing to stop hard angles causing displacement breakages. This is in fact one of TG2s unique and great features. In other apps you end up with displacement artifact hell. Actually, that leads me to think about dot product again. There are probably a host of uses for that in displacements by manipulating the angles. I was thinking purely in terms of colour before. Thinking back, I'm sure I have files where I've used dot product to actually smooth angles.
Could it be usefull in making a step edge more verticle? By say using the ramp between each step to lateral displace oppposite to it's x,z normal?
I don't know but I'm going to take a look at using the modulo again today to see what else it could do.
We have a way to create large smoothed steps. Even if my file isn't the best graph (there could be a better solution) what it does is what I was after. You can manually move large steps around to create cliffs, mesas or whatever. What we are lacking is a function to create steps similar to strata and outcrops but without the hard edges. Even strata and outcrops has a repeat so in some circumstances you can see that. Ideally we could create strata if the fractal allowed us to change it's position to altitude like we can with the noise functions. However, this doesn't really create clear steps anyway since it uses the fractals profile if you displace it.
Actually, it's not even so much that we have edges with the strata and outcrops but it's the facteted nature of it. This creates a certain kind of hard rock look which is cool but I'd like to have more smoothness especially to create slightly larger steps.
You can already shift a PF with the Warp shader. I've used this to shift a PF around a point, to get an average value, which can be used to smooth out the slope of a road along it's length and it's width.
I'm still not sure on how to get random spacing between steps, I'm thinking an altitude into a voronoi cell/difference, but I've run into problems with that approach before. Goms mentioned sampling an offset sine curve in an old thread, which might be the way to go. I shall have a play with some ideas tonight.
It's easy enough to distort the modulo. I've tried that and it's quite cool. You just mix two values for it with a Perlin. I've tried various altitude methods but that seems to be no go.
I'm not sure what you mean about shifting the powerfractal with a warp shader. I know you can do stuff in this direction as far as moving the powerfractal around but the biggest fundamental flaw of TG2 is not being able to simply change the position coordinates that the powerfractal uses. It makes proper masking impossible, for example. This limitation has to actually be changed. Matt has suggested that he will do this. It would certainly make TG2 unviable if it isn't changed.
You can easily move the texture space within a chain, thus repositioning the origin. Plug a PF into a warp shader's shader input, then use a redirect conected to the warper input. Then plug a displacement shader into each direction input you want to use on the redirect, then your values into the displacement shaders. This will shift the PF around what values you drive the displacement shaders, wether that's a static scaler, or a modulated value. I've used this technique to twist PFs into cone shapes for a hurricane, and sample another PF by offsetting the origin to determine if I create a sphere in a 3d grid location in one of my cloud experiments.
It's a bit convoluted, but there are plans, like you mentioned, to make this easier. But it can be done right now.
Quote from: efflux on March 14, 2012, 12:50:32 PM
You just mix two values for it with a Perlin.
This is something I should use more often. That sounds like a good idea.
Yeah, it's crazy sometimes the way you just hook something up that is simple and works. Of course there are lots of ways you could essentially distort the modula steps, after they were actually created for example so this is not really a unique thing. What we want is for it to change step size vertically.
As for the positions thing. I can see how you can move that around. Matt even says that there is a way to change surface position to another position but that would need to be say, geometry position. Not really sure how you'd do that but it's really not viable anyway. You're node setups would be ludicrously complicated.
This is what you get when you use a perlin to mix the modulo input. You can radically distort it up way beyond this.
Nice.
I went a little harder with my displacements.
That's looking very cool. I think this modulo angle is going to yield some great results. Even just shifting it around a bit makes it more natural. I was thinking that subtracting a voronoi 3D A on steep slopes like I was doing in the terrain altitude blend thread would create vertical rock like features which blended with the horizontal modulo steps would create very nice chopped up rocks.
Yes, I think it would.
The displacements I used were a bit extreme, and I didn't have final normal switched on for my rock population, so there's some exploding stones in the below pics. Shame, as there were some interesting structures.
Nice results but I didn't go further with the modulo idea because it would mean going in a very different direction with the basic planet file on here. It's not suitable to integrate the modulo but it definitely needs further exploration.
I wanted to totally push the limits of some things I've been doing with terrain so I used the terrain in the first post here and brought in subtracted voronoi limiting it to the cliffs with altitude terrain blending. This is about as far as TG2 can go along these general lines. This is really pushing the limits with only terrain. Very steep cliffs as well. I'm surprised it even looks this good. No surface displacements at all so in essence it doesn't even need to be this extreme. Surface displacements would do a lot of work but I'm into pushing the limits. It can be improved on and optimised to get cleaner results. Probably better graphs as well.
In the other thread, where I used voronoi, I actually negatively displaced it. Here I subtracted it from the main terrain via a merge shader just to be accurate about it. That should be a no brainer for people to try though. It's also several voronoi of different sizes simply chained together.
I might tweak around and improve things a bit and post more renders but I'm done with TG2 for a while I think.
Very nice stuff, guys. Before you go, efflux, a nice tgd or tgc would be very informative ;D I do hope you'll continue with your explorations in the blue field.
Hi Hetzen, I've sorted out the stepping but I cannot get a smooth transition :-[ any help even a screen shot would help. Thanks efflux for starting these threads very inspirational! Mick
I'm liking that last image a lot Efflux. Loads of potential.
Here's a screen shot of a stripped out version I was using Mick, with explainations if anybody else is interested.
EDIT Nutz! There should be a divide by the 'difference scalar 01' before the 'smooth step', to bring the ramp peak value back up to 1
Cheers Hetzen, I understand now - WHOPEEEEE!!!
Thanks Hetzen.
When I read the messages I though I'd have to sort out another tgd. Your graph is basically the same as mine. I've really had enough and need a break. I spent ages on some of the images you see in my WIP thread.
Dune.
I'm not sure if you mean a tgd of what Hetzen has come up with here on the other thing I was doing. Hetzen has solved the other modulo thing with a graph screenshot. My last image actually has a bad graph idea that I realise now. What you often see is me testing things but I don't necessarily give a tgd because sometimes I find problems.
The correct way to do the last picture I posted here (not the way I actually did it) is to use the terrain I posted at the start of this thread and incorporate that terrain in the second file of the Terrain Altitude Blend thread. That's it. You'll have to tweak it around to get a desired result. I have no file as I just described. I used subtract instead of negative displacement so it's not the best solution. You may find that this doesn't get you the exact same look (I haven't experimented totally with it) but the last picture here creates too much calculations and isn't a nice clean method. I won't use that again.
Now I know a way to further what I did in that last image to make it cleaner.
The problem is guys. This just goes on and on. I have dozens of ideas where there isn't even any work done yet.
I'm constructing a full blue node terrain stepper. The red node version is too slow. This will require a mass more blue nodes though.
Stepped terrain version 2 will be coming. This is WAY better than the last file. It's all blue nodes. Has to done this way.
I'm just going to look into the possibility of building in some other functions before I post it.
OK, there are problems with this. Basically you can remove the red displacement nodes except the last and replace them with scalars chaining the thing in the same way. What scalars? Well this is the problem. There are a hundred and one solutions and how they are interacting with the final displacement shader is totally relevant. Every different solution gets different results so I can't see how to build a do it all graph because one that works in one way will create a problem in another way.
I've built lots of stepped terrains and every single terrain has a different shape and a different graph depending on what I need to do. I have no idea what kind of steps people are going to create. Every one has a totally different graph some with different profiles of steps in places.
Amazing work. This looks just like the SW US. Keep it up!
digitalis99.
You highlighted the problem I just mentioned. I can build a stepped terrain that is better than that. Looks exactly like the kind of mesas you are talking about but it won't be good for other things.
This is why we need a curve graph!
Different need = different tool
I'm sure you'll come up with something.
OK guys. I've come up with this simple concoction which allows a lot of variety except changing step profiles. Changing the step profiles creates a lot of extra hassle. Basically each step has a multiplier (or divider). This works in conjunction with the final red node displacer. The severeness of the steps depends on what relationships of values you use between the multipliers and the final displacement node value. There are no hard rules as to what values to use except what looks right.
If this one is deemed good I'll put it in the first post so people find it more easily.
EDIT:
The file is now on the first post of the thread.
It might not actually be that clear why I made the graph this way.
Set all the multipliers to 0. Increase the displacement node value to get a bit more height then render. Set the multipliers to 0.05 and render again and so forth. I tried to give a bit of versatilty without it getting complex.
Thanks for all this, efflux!
My version of a stepped Perlin (scale 1000, displacement 800)
That's interesting Mogn.
What we should do is make specific stepping graphs for specific styles of terrain based on still having the ranges of steps changable but design the profiles of each step instead of always using smooth steps. Since we don't have a curve graph, this is the only way. For example, the monument valley buttes. You'd want another profile for the deposition area. The general arrangement would be tweakable but each step region would be a specific part of a specific terrain. For example, I found that often a sharp edge on the top ridge works rather than using smooth step there. This is one place where sharp edges work because it's a hard rock outcrop.
Next test render.
I think I'll delete the file in the first post and move the newer stepped terrain there.
This is another blending of the various methods. Much better technique this time than in the cliff face in that previous render. It's different but that one was not efficient. Stepped terrain and voronoi negative displacement with altitude blending of terrain. What I'll do is sort out the more complex voronoi thing and put that in the other thread. You want to start with the steps first anyway then add in the other stuff. The files will be self explanatory as to where to put stuff.
Mogn, I just did the add for the cascaded voronoi this time. No need really for extra multiplications to get different displacements no matter if blue nodes or red nodes. This is rendering fast now. I'll take a look at what you did with those blue nodes. Those kinds of soft undulations are cool.
Next phase will be to split this off into Bryce Canyon and Monument Valley. Bryce Canyon is possible but it might have to wait.
Speaking of Bryce Canyon, i found an interesting perspective of such features in that area.
Bryce Canyon will be possible. I just didn't get there as soon as I thought I would. TG2 can handle the surfaces of Bryce Canyon no problem. As for the terrain, we need stepped terrain and we need a deposition effect. The towers will require a lot of tweakery with shapes that cut into the steps. The last image I did shows slightly more of this cutting. It's still a step but it's cut into. I have another file I did a while back that goes slightly more in this direction but other than that part of the technique it doesn't look very cool so I never posted a render.
The file at the start of this thread is deleted and replaced with the newer one. I didn't want that to remain there because it was quite slow to render.
The interesting thing about Bryce is the deep cut erosion is what creates the tall features so it needs to be iterated a few times which looks like you are close to that already then doing the sediment talus at there bases but they have to match up to look as of it is flowing from the cuts themselves. Way beyond me. Everything else, the horizontal strata types and larger scale displacement along the eroded mounds would be easy to do. These are very interesting tests and can't wait to see the end results. These can act as a springboard to make some neat features for Canyonlands National Park in Utah as well. Arches is a whole other story. ;)
Quote from: efflux on March 16, 2012, 10:57:23 AM
That's interesting Mogn.
What we should do is make specific stepping graphs for specific styles of terrain based on still having the ranges of steps changable but design the profiles of each step instead of always using smooth steps. Since we don't have a curve graph, this is the only way. For example, the monument valley buttes. You'd want another profile for the deposition area. The general arrangement would be tweakable but each step region would be a specific part of a specific terrain. For example, I found that often a sharp edge on the top ridge works rather than using smooth step there. This is one place where sharp edges work because it's a hard rock outcrop.
The reason that the soft step works giving the soft looks, is that the slope in the endpoints are zero.
Your remark about missing a curve graph give me an idea:
Replace the soft step with a spine curve defined by seven values, the two first values and the two last values taken from the Perlin, and the 3 middle values
from another fractal e.g. the 3 components of a voronoi vector. This will give a soft but varied landscape.
Those ideas sound interesting.
I haven't actually looked at your graph yet. The main reason for this is that I'm taking a break for now with TG2. I'll probably look back at this stuff some time in the near future.
Here's a small variation on the final steps file at the start of the thread. Instead of using a constant value to create the step heights, this file uses a perlin noise so that the steps don't occur at the same level of height. It makes for a distorted stepping like many terrains in the real world. Each step could be treated in all sorts of different ways. The constant for the step heights in the first file could be distorted.
Here's another variation. Each step is treated differently. This gives more variety.
Here's another variation. I just tweaked some parameters in the Steps 10 file. Instead of those regular steps we can get subtle step height changes.
Here's that last scene slightly tweaked with some simple colour and water added. The first step is around the shore area. Sometimes going beneath the water. This kind of thing makes the water blend in as if it's reacting with the terrain. Eroding it, depositing or whatever.
Uuh, it looks fantastic!
As Noobie i would love to get the chance to analize the last .tgd scene with the water and stuff..
Could/would you share this file too? :D
*Cheers
Alex
I'll be back on my Windows system soon so I'll get that last file but I'll remove some of the rock colouring, there is something else going on there. I want to keep the files simple. It's really just the previous file with one step slightly altered, then water and colour added. One thing that is making my renders look bright and crisply lit (at least when I get to adding any colour) is some post processing in an app called Lightzone. It has some fantastic algorithms for relighting scenes. Especially great for landscapes where lighting differences are extreme. Same goes for TG2 - if you are replicating the real world of cameras and photos.
You will see that the water roughness has very small values. This all depends on the scale of your scene (I'm not even sure exactly what kind of scale things are in these scenes) but most of the time water only needs very small values of roughness to be effective unless it is an open ocean.
Here is that last tgd but with only the one colouring fractal. Nothing special really. I just put water in to try to show how it could relate to a step. It's not a set up I would really design. The step is quite steep.
Thank you very much efflux. Many thoughts here well taken. Hope to be combining with some of my stuff. You deliver some great inspirations here.
Quote from: efflux on November 10, 2012, 02:12:54 PM
Here is that last tgd but with only the one colouring fractal. Nothing special really. I just put water in to try to show how it could relate to a step. It's not a set up I would really design. The step is quite steep.
THX m8!
For me as nub is everything what is looking good/inspiring something "special". :D
It´s often informative, to compare tgd´s with different approaches/outcomes of several "TG-Veterans".
Reverse-engineering is hard, but fun and enlightening. But shocking too, sometimes, if you just try to find a "simple and fast" solution for whatever.. :P).
o/
Alex
Alex,
Often the solutions are simple but since this way of creating 3D doesn't have a huge history and backlog of methods to find out about, it can be hard to find these solutions. The other problem is that often images may look like they have some extremely fancy graph work but in fact it's just that the creator has tastefully chosen ways to get a convincing result. It can be very hard to just get a rock colour right never mind all the other considerations.
Just another point. I never start from a point of view of creating a "realistic" scene. Some of my older stuff is crazy. I test what I can do with the program. Sometimes I find something I recognise will get a result that looks like a real world earth environment so I use it for that. That's why my more recent stuff appears to be heading in a more realistic direction. I have some gripes with how TG2 works because there are some things I know could be done to get realistic results but the app won't let you do it.
.
Here's two more images along the lines I'm working on here. Both stepped. I'm on Linux at the moment so can't give specifics or files but if there is anything new from what I've already posted I'll post files. Steps 14 shows you how this is more subtle with the distortions. The hill on the right has a cliff but it's uneven. Natural looking. Dust 2 is a Perlin terrain that is inverted. Just one method to change the forms. Smoother tops on hills and rocky cliffs towards shore. Ther stepping is quite fundamental to the forms here even although not immediately obvious. This is meant for further tweaking to create beaches and more water area hence the "Dust" name. It's meant for a new planet concept. The "water" is not a water shader as you can see. It's a second planet that will eventually get a water shader but no point as this stage. Water just increases render times.
Thanks very much efflux for the extensive sharing of this subject !
Will love studying this in the next few weeks. Much appreciated !
Paul