Rendering water in TG

[monks]

I've posted in the TG boards on this here:
http://forums.planetside.co.uk/index.php?topic=1864.0

...maybe someone will have a bright idea. :) I know you guys have tried this, maybe if yourself Johannes or Criss could provide a link to some files and more explanation on the difficulties? ME-DEM will happily host any files. I think this is important and it is actually breaking new ground, but everyone wants to see renders of this stuff and I'm sure there must be someone in the TG community to could solve this very quickly...!

monks

[Jules Verne]

I'm not sure I follow you. Water with height? You mean just different levels of water, say a lake in some mountains, an ocean far below and so on? Or frozen water thats built up over centuries where it used to flow down mountain sides?

Put one height map terrain inside another, one with a water surface one regular. Or use flat polygons with a water effect on the surface. Or volumetric cubes if you plan on doing things underwater?

I'm probably missing the point aren't I? :)

[cajomi]

I have tested it in Carrara.
As long as TG2 has no transparency you can simply use the water filled modus and shade the water parts with the river mask.

[Criss]

Well, i have stopped using Terragen 2 for a while now as it takes to many steps to what was once simple in the old Terragen. Case being the heightfield imports, texture maps and masking of course so for me its a huge headache with something that should only take a couple of steps rather then several unnecessary steps. For the time being it is very easy in Vue and Carrara but those programs have there limitations as well in terms of realism and how Vue imports Heightfields poorly in terms of resolution. Carrara just has poor skies and object instancing handling.

The next demonstration image i am working on at the moment will be done in old Terragen.

[Criss]

I can of course provide some cool masks and a Terragen file. I will be out over the weekend but i can make up a very nice set of files from working with the river tool etc. so just give me the latest until monday and perhaps someone can play around with those masks and the heightfield in Terragen 2 for the heck of it.

[monks]

Jules, it's been a real pain to get any of GeoControl's (and Geomorphers) water to render with all the bells and whistles. To be frank, I'm not sure exactly what the details are, but I posted over there in the hope that we could sort something out. In any case, I think the old addage applies: 'if you want a job doing properly, do it yourself' ;)

Righteo Criss, they might be useful. I guess this a problem that I'll have to have to work through at some point

monks

[Jules Verne]

I've always had a habit of doing things the tried and tested ways rather than specially made methods of doing things. So I would probably do it in one of the following ways.

For general appearance, from a distance and not requiring close up scrutiny.
If I couldn't get an output of the rivers on their own as a B&W alpha channel. I'd take the heightmap over to either ZBrush or EarthSculptor and draw in the rivers to create by hand to create the B&W image. Take it into an application like Lightwave or Softimage and use that image to create a watery texture without it effecting the rest of the terrain.

If I was under a deadline. I may even just draw the rivers/lakes in without putting much thought into terrain. Depending on the situation a client may not notice the terrain isn't 100% perfectly accurate. So long as it looks ok they're generally happy enough hehe (don't tell em that lol)

If the water is required to be viewed close up, or interacted with during an animation
I'd decide what will be "hero" water and what wont be. The hero water would be modeled as separate objects, while the rest would be similar to the above method.

[b][For just a single image/b]
Any water not close to the viewer, I'd paint into the image later by hand. anything close I'd do as a separate object and place it into the terrain for rendering with it. If that wasn't possible. I'd fake it, render some water surface with the terrain around it for reflection, output the water surface on its own and merge it carefully into the final image, then clean it up.


I guess it depends really on what it is you're aiming for. An animation, such as a flyby or just background effect, I'd probably fake a lot of it. A single image, I'd probably put it together in post rather than attempt to do it all in one program.


But yeah, these are long winded ways of doing things, which isn't for everyone. I just tend to prefer to trust myself more than software. I don't even like computers, the irony eh? :)

-

I wrote out a bunch of different methods, which I realize likely isn't going to help you one bit, but on the off chance someone may find them useful I'll leave them in the post below.


But first could you post examples if possible? I'm still a bit unsure so I don't know if any of my suggestions will help. Plus they may not even work in Terragen as I've barely used it in years.

At the moment, from what I can understand. A workable solution for water to appear in streams, rivers and lakes of any altitude, which look fine from a distance but wouldn't hold up to closeup examination. Would simply be a grayscale image where white is water, black is nothing. Which you'd then use as an alpha channel or mask, over the entire terrain, to prevent the watery textures from showing through anything but the area's you want it to appear.

On the link in the original post it mentions height maps, if you have access to photoshop/paint shop pro. You can change the contrast/brightness of the image to flatten the heightmap and turn it into a straight mask. Or change the threshold around, creating variations of the same mask, but allowing different textures through at different elevations (allowing for simple waterfall effects, rushing water and calm still water all at once)

If you have access to something like Softimage, Maya, Lightwave etc. The above solution would definitely work. Also with Maya you could use paint effects to draw in the water, or in any of them use vertex/weight maps as a mask, though those will depend on how detailed the terrain is as they rely on the vertex's in the model for clarity.

Like I say, these aren't likely to be much use in Terragen, but if you're using others then they might be an option. As for skies, you could always use photo's, it's generally not really minded when someone does since the work required to generate realistic clouds digitally is more trouble than its worth half the time. :)

[monks]

Thanks for the info Jules- I think you are probably a long way in advance of me. ;)

I do have access to Maya actually through a friend but that's not really an ideal workflow for me. I don't plan on acquiring yet another app, that'll probably never be used- as much as I'd like to!

My skills at the moment lie more in terrain building as opposed to rendering. Everything I've learned in the last couple of years has been a result of a terrain modelling project I'm involved with : Middle Earth DEM Project (http://www.me-dem.org/component/option,com_joomlaboard/Itemid,/func,view/catid,4/id,1284/#1284). We've still not really gotten around to full production renders yet...!...so that's why things get a bit fuzzy shall we say at that end.

We're very much looking to see terrain modelling/visualisation advance; something which I believe is happening at a pace with apps like GeoControl et al. Check out the Terrain Summit (http://terrain.cg-arts.org/forum/index.php) forums, both for like minded chatter and to promote your softwares.

You sound like a knowledgeable bod to have around. I've tried your Earth Sculptor and I like the navigation in it. The processing of the water effects is arguably better spent on other things, but I can see that the app is still under development; it does make a pleasant change though to have pretty visual effects like that in a terrain editor and I'm all for game technology being used in terrain editing software wherever possible, especially lods.

What size is the demo terrain, 512?

monks

[Jules Verne]

Thanks for the info Jules- I think you are probably a long way in advance of me. ;)

I do have access to Maya actually through a friend but that's not really an ideal workflow for me. I don't plan on acquiring yet another app, that'll probably never be used- as much as I'd like to!

My skills at the moment lie more in terrain building as opposed to rendering. Everything I've learned in the last couple of years has been a result of a terrain modelling project I'm involved with : Middle Earth DEM Project (http://www.me-dem.org/component/option,com_joomlaboard/Itemid,/func,view/catid,4/id,1284/#1284). We've still not really gotten around to full production renders yet...!...so that's why things get a bit fuzzy shall we say at that end.

We're very much looking to see terrain modelling/visualisation advance; something which I believe is happening at a pace with apps like GeoControl et al. Check out the Terrain Summit (http://terrain.cg-arts.org/forum/index.php) forums, both for like minded chatter and to promote your softwares.

You sound like a knowledgeable bod to have around. I've tried your Earth Sculptor and I like the navigation in it. The processing of the water effects is arguably better spent on other things, but I can see that the app is still under development; it does make a pleasant change though to have pretty visual effects like that in a terrain editor and I'm all for game technology being used in terrain editing software wherever possible, especially lods.

What size is the demo terrain, 512?

monks

Eeek!

I'm not the guy who made EarthSculptor! *blushes* I wish I was that talented! lol! :) I can't believe I gave that impression! God now I'm embarrassed hahaha. I've no idea the size of the terrain in his demo, I just linked to it to since it showed clearer than I was explaining, the same method of displaying the terrain in realtime. Though you can run Earthsculptor with a 4096x4096 and only medium LOD, with everything textured, pretty far view distance and fog at fullscreen with little to no slowdown.

I just liked the method that was used so decided to use it in Gaia (which is mine hehe). I've just been working on a hobby of mine the past few months. I came from the art side of things and am slowly learning to code. My real area of expertise would be the art side of things, but then not so much in things like Vue or TG since I've mainly stayed with apps like Lightwave and Softimage.

Saying that, Earthsculptor does work nicely when used with GeoControl's output, but it can be a bit flaky at times with larger terrains. But for fine tuning a terrain after doing the bulk of it in GC its probably a good choice, also its free haha.

So while I might not be able to help out much on the programming side of things since I'm learning as I go. Art side, tricks and ways to achieve things easier I can probably help out on, learn't a lot from friends who worked on a lot of big movies in the past and would pass on the things they learnt under tight deadlines to the rest of us, and I'm using a lot of those pre-rendering tricks in my game to make what I plan to do more plausible and realistic.

So if you've any questions relating to that side of things, I should be able to help out. But yeah, I'm not the guy who made EarthSculptor, I wish I was, I'd be so much further ahead with my project then hehehe.

[monks]

hehehe -oops, well Gaia looks pretty nice too and if you're using the lod technique described there, it should be pretty capable. I can see you've made lots of optimisations with tradeoffs. It's a renderer, not an editor?


This final image should I hope show my need for being able to tile the terrain and generate higher quality tiles from that one master terrain.

This is exactly the way we're looking to create the terrain at ME-DEM :) That is, using a 'procedural zoom' on a master map and then tiling. It makes a LOT of sense.

monks

[Jules Verne]

hehehe -oops, well Gaia looks pretty nice too and if you're using the lod technique described there, it should be pretty capable. I can see you've made lots of optimisations with tradeoffs. It's a renderer, not an editor?

Thanks :) Got a long way to go with it yet though. But the basics are there, and the way I'm doing it, which I suppose is down to me being an artist more than a programmer, the real meat of it will be in the artwork. It would be great to have something with the quality of say the Unreal Engine and be able to wizz around hundreds of kilometers of landscape with billions of plants and tree's and everything be physics based and blowupable (if thats even a word hehe) and all with realtime lighting. But something like that just isn't going to happen for a very long time. I realized pretty early on that while something neat like a physics engine, letting people move just about anything in world is pretty cool and fine for single player games. For an online game it ultimately causes more problems than its worth. So deciding I wasn't going to have that in has helped a lot.

It also doesn't _need_ super realistic lighting, which let me use the method I decided on, where all the lighting is baked into textures using global illumination rather than a direct point of light, which would give a very diffuse effect, which except for some rare occasions is how the outdoors look in the real world. Doing it that way then had the knock-on effect of making day and night cycles easier. Then with everything done in tiles it cuts down what needs to be loaded at a given time. Nothing new in it really, just idea's and methods from various other engines thrown together.

The big downside of course is a huge number of textures and duplicate objects for the static ones (since each and every one will use a unique texture) which will make a download of the entire thing pretty darn big. But I figured in this day and age, the kind of people who play online games will generally have broadband and be used to large downloads (especially if they play World of Warcraft hehehe)

No its not an editor, I thought about it at one point but yeah, like I said in one of my earlier posts, why reinvent the wheel when there's editors like GeoControl that can do that side of things. Means I can spend more time on other area's. I know the same could be said about the engine itself, when there's ones out there like Torque. But they didn't work the way I wanted them to, which was to specifically use a pre-render application as the entire editor. A personal challenge to myself I guess.

This is exactly the way we're looking to create the terrain at ME-DEM :) That is, using a 'procedural zoom' on a master map and then tiling. It makes a LOT of sense.

monksIt's perfect isn't it for this sort of thing? :) I was like OOOOOOH when I realized it could do it lol. Originally I also looked at World Machine for doing this sort of thing, and as realistic as the output is with that software, it suffers from the repeating procedural effect, so while you can fly around in that and cut up sections to suit, and zoom in all you want. The terrain will ultimately repeat as you go. Which was a shame. Same thing with Mojoworld, and even creating terrain directly in a pre-rendered app.

I know we'd get more control doing things by hand, but our projects, things are just TOO big to be doing by hand aren't they? :)

How big is your entire project btw? I'm a huge fan of LotR, in fact it was one of the things to inspire me when I first started this years ago at school (its changed a lot in the past 15 years LOL!) yeah 15 years its taken me to finally get to this point! I would imagine the entire surface of middle earth would be on a par with the entire size of Europe? That's huge hehe, but if GeoControl 2 will allow for tiling and seamless tiles at that.. Wouldn't that in theory allow you to build the whole of middle earth within a single project? Wow.. hehe I can't help but grin thinking how much fun that would be hehehe.

For mine, the terrain I would like to create that way. Build the entire location in a single image, which in itself would be useless to deform each tile (I think it would be something like 8x8 pixels per 1km² That would be just ugly when your down there on the ground) So the single image terrain wouldn't be used by itself, except for maybe as a base for a hand drawn map perhaps. I would imagine in my case, the full single image map would have really over the top filter settings, it will I think look terrible as a single map like that. But the much smaller tiled area's would then have the details they need, and from their point of view, it should (in theory) look pretty nice. The tiling feature would be what would provide the actual heightmap data for deforming the grids. Cut a tile every few pixels, zoom in and save out a much higher resolution tile. Then I've got two options, use the app I'm using for the editor to fine tune each tile, or Earthsculptor, and walk around checking there's no problems. Big job, lot of work, but fun!

Without tiles. Assuming each grid was deformed with a 1024x1024 heightmap. I'd have been looking at a full single heightmap of 524288x524288... *faints* I don't even know how big a file that would be! LOL. Photoshop only lets me go up to 30000x30000 and that would be 2.5GB :O and thats in 8bit too, and I'd need it in 16bit, so yeah tiles are the only way I'm ever going to do this. :).... Unless I end up just doing it all by hand, which I doubt I will at this point hehe.

So far mine is probably only going to be around 512km² or there abouts, depending on how things go. The engine can quite happily handle any size, but I'll be editing all that terrain by hand, by that I mean placing everything (except grass, that'll be done in the engine like in most games which display grass), so I can't go too big or my wrists will breaks lol.

Do you intend to also add buildings, plants etc. in yours? Or is it going to be just DEM files? It would as you probably realize, be an almost impossible job to add everything onto those maps.. But damn, that would be so much fun to do that! :D A fully complete and detailed fly through of the entire middle earth. Wow! :)

[monks]

The terrain will ultimately repeat as you go. Which was a shame. Same thing with Mojoworld, and even creating terrain directly in a pre-rendered app.

Yes, that's it in a nutshell. You have to have a mastermap- what we refer to at the ME-DEM Proj as a 'base dem'- which is developed from raster maps. We started this a couple of years ago and at that time you had only one option for dealing with uber large files of any kind: Photoshop. So that means you're starting out in the raster world.
Then there's the problem of 8 bit vs 16 bit greyscales. Back then we only had the 8 bit option, so we decided to go with raster contours -> vectorisation -> elevation tagging. That workflow went a more GIS route (and not for everyone)- and the GIS industry generally has good support for large files (for obvious reasons) so that helped.
I see the master map as like the seed from which your world grows- the procedurals are guided by the dna you lay down in the seed.

I heard something like 80% of resources are spent on content creation in games. Don't now how accurate that is, but if it ain't that now, it will be soon.
Horses for courses. I think the above outline is a good compromise between procedurals and human input. Another behefit from the proc zoom is the fact that you can redo/improve the terrain (and parts of it) with a lot of impunity. So really we look at our terrain data as more of a persistent data set- changing/improving over time- in a way very much like the databases of real world dems out there on servers.
We've talked on the GC boards about ways to guide the filters across the map and I think we need both control over where the filters are applied (ie selections) but also how those filters change over the course of the zoom. Having the same filters over the entire zoom is not entirely satisfactory- beyond a certain zoom threshold. That's a new problem and I think Johannes mentioned the possibility of new types of filters.
You can see that here:
http://www.me-dem.org/content/view/56/33/



How big is your entire project btw? I'm a huge fan of LotR, in fact it was one of the things to inspire me when I first started this years ago at school

Excatly, if you can't have fun doing it, why do it! :) Yep, we're hoping to build the whole of Middle Earth- starting with the north west -ie Europe. The terrain is going to be georefferenced as well -so it will map to a globe such as Google Earth or World Wind, etc. THat'd one of the visualisation platforms we'd ultimately like to see the dem and maps on. We're using Earth as our template since that's in fact what Tolkien based Middle Earth on- (M-E was meant to be Earth but in an alternate history- roughly 6000 years ago). The terrain is open for any non-profit/commercial use to all ME-DEM members.

For mine, the terrain I would like to create that way. Build the entire location in a single image, which in itself would be useless to deform each tile (I think it would be something like 8x8 pixels per 1km² That would be just ugly when your down there on the ground) So the single image terrain wouldn't be used by itself, except for maybe as a base for a hand drawn map perhaps. I would imagine in my case, the full single image map would have really over the top filter settings, it will I think look terrible as a single map like that.

Yes, like caricatured: here be mountains, here be deserts. I think it'll depend on how much contol we have over the filter-zoom process.


In terms of convincing geography, my advice would be to emphasise 2 features in your base map: upland areas (areas from where rivers will drain) and rivers. You'll no doubt need fewer rivers than say ME-DEM but I think that will give you a good way to orient yourself when the map starts getting crazy huge! -but more importantly (imo) it will give you believable terrain- it depends how important that is to you. If you get the big features right, GeoControl will do (most of) the rest for you.

Without tiles. Assuming each grid was deformed with a 1024x1024 heightmap. I'd have been looking at a full single heightmap of 524288x524288... *faints*

Soon we could (if we have the storage/ time) be getting into the realms of terrabytes of data. The ME-DEM base map is 20K x 20K- that's really about as far as I could push the hardware I was using back then. The plan was to upscale by factors of 10 in 2 stages. So 200,000 was the immediate goal. Now that's gone by the board as a necessary plan but we'll stil probably work in powers of 10.
I'd like to get the base map to 40K *if possible*. Then that map will serve as a base camp from where we can zoom.

So far mine is probably only going to be around 512km² or there abouts, depending on how things go. The engine can quite happily handle any size, but I'll be editing all that terrain by hand,

Morrowind was a good size. I think once you can get travel boredom, you have a believable world (though not necessarily enjoyable experience). I think the best way to guage it is to now how long it takes in real time to travel across the map. Did you ever get a look of the mmorpg Dark and Light?; they've got some great terrain in that.

Do you intend to also add buildings, plants etc. in yours? Or is it going to be just DEM files?

Hehe- yep, that's where programmers come in very handy!! If you can write procedural solutions/ scripts to placing objects in your world. then you're potentially onto a real winner. Obviosuly you'd want to place any objects central to the game/plot by hand.
Yes, we're going to add everything we humanly can- but we don't have any deadlines which is just as well for an absurdly ambitious plan :) Like I say we hope that the data (dems. 3D models, maps, etc) will stick around and be (re)used by people for all kinds of things: games, renders, online maps, etc. M-E is such a perennial favourite!
We have a developer who is working on a procedural texture generator. (Incidentally he's also building a huge world like yourself). Basically this says what type of enviroments are where based on temperature and rainfall (and we hope latitude). It spits out a 'base' map which functions rather like the base dem. Then it upsamples and stitches photos together. You could use colourmap/texture output by GeoControl or any terrain app, it's just a case of programming the compositor to know which input pixel colour corresponds to which habitat. It then goes off and finds suitable photos and stitches them together....that's the theory anway!
I'm starting to plug things here- so feel free to continue the convo along these lines on the two boards linked. :) Why not post your progress to the community here as these guys?:http://terrain.cg-arts.org/forum/index.php?board=22.0

monks

[Jules Verne]

Sorry it's taken so long to reply, had a few girlfriend's family issues straight out of some badly written comedy show haha.

Yes, that's it in a nutshell. You have to have a mastermap- what we refer to at the ME-DEM Proj as a 'base dem'- which is developed from raster maps. We started this a couple of years ago and at that time you had only one option for dealing with uber large files of any kind: Photoshop. So that means you're starting out in the raster world.
Then there's the problem of 8 bit vs 16 bit greyscales. Back then we only had the 8 bit option, so we decided to go with raster contours -> vectorisation -> elevation tagging. That workflow went a more GIS route (and not for everyone)- and the GIS industry generally has good support for large files (for obvious reasons) so that helped.
I see the master map as like the seed from which your world grows- the procedurals are guided by the dna you lay down in the seed.

There's another option, if all digital methods fail... And its a crazy one that just came to me. But you could always make a scaled clay model and have that scanned hehe. It just popped in my head when I was reading the above and how I doubt I'd get anything done if I couldn't atleast see or feel the terrain properly, the whole contour thing is a bit beyond me I think. :)

I heard something like 80% of resources are spent on content creation in games. Don't now how accurate that is, but if it ain't that now, it will be soon.Yeah it's definately around that figure I would think. Players want more and more and these days just a small game is getting to have a movie like production. i say bring back the old days where people made games in their bedroom during the weekend lol.

Horses for courses. I think the above outline is a good compromise between procedurals and human input. Another behefit from the proc zoom is the fact that you can redo/improve the terrain (and parts of it) with a lot of impunity. So really we look at our terrain data as more of a persistent data set- changing/improving over time- in a way very much like the databases of real world dems out there on servers.
We've talked on the GC boards about ways to guide the filters across the map and I think we need both control over where the filters are applied (ie selections) but also how those filters change over the course of the zoom. Having the same filters over the entire zoom is not entirely satisfactory- beyond a certain zoom threshold. That's a new problem and I think Johannes mentioned the possibility of new types of filters.
You can see that here:
http://www.me-dem.org/content/view/56/33/
Layers should do the trick there wouldn't they? Atleast in the way I'm assuming they work (and have been using them) each layer representing a certain part of the terrain. For example, layer one, your base layer would be for the underwater area's, giving that the underwater rock appearance, maybe another layer on top of that just to mix up the styles. Then you'd have your base ground layer, probably just a generic coastline and the actual bulk of the terrain. Then further layers would each handle different things, dunes, meadows, small hills, large hills, rocky mountains, snowy area's and so on. That's how I've been using layers since I got GC the other day, each one has a different set of filters that can be changed separately from each other, then everything comes together in the project tab. Means you can edit various area's to your hearts content without effecting other parts. Decide to add a beach or a range of mountains, add another layer and try it out, don't like it, clear the layer. That was why I asked in another thread for an onion skin effect, so I could better see where I'm placing things.. Course I'm thinking maybe I'm not using layers in the way they were intended? hehe oops. (But it does work well like that heh)

Another handy side effect of using layers like that, is you can zoom in and change terrain resolution all you want and the filters wont effect the entire landscape, only those area's they should. I guess you could say it is a selection of sorts really. I suppose with the overlay feature, the onion skin thing could work that way perhaps? At the very least giving an idea where things should be going, so you don't suddenly place mountains out in the middle of the ocean (which I have a few times oops)

[Jules Verne]

Exactly, if you can't have fun doing it, why do it! Yep, we're hoping to build the whole of Middle Earth- starting with the north west -ie Europe. The terrain is going to be georefferenced as well -so it will map to a globe such as Google Earth or World Wind, etc. THat'd one of the visualisation platforms we'd ultimately like to see the dem and maps on. We're using Earth as our template since that's in fact what Tolkien based Middle Earth on- (M-E was meant to be Earth but in an alternate history- roughly 6000 years ago). The terrain is open for any non-profit/commercial use to all ME-DEM members.

You may have seen this..
http://www.uesp.net/oblivion/map/obmap.shtml
The entire map in Oblivion, displayed in a google map interface. I briefly looked at how they put it together and it too uses tiles in a very similar way to how you seem to be. So it could end up being pretty simple to make a google maps version of that for your project. Showing whats done and what needs to be done, aswell as links to various area's, showing off proper renders of different locations. And you mentioned Google Earth, it would stand to reason both of those would use similar methods of displaying data.

I've thought about using the same thing myself, as a private map while I work, so I can better plan where things are going to be. Right now I use a private wiki for setting things up, much better than using lots of text files like I used to haha.


Yes, like caricatured: here be mountains, here be deserts. I think it'll depend on how much contol we have over the filter-zoom process.

Exactly, from testing it I get the impression it could definitely work too. Though I have noticed in the alpha that the isolines don't follow the zooms, took me forever to work out why there were familiar markings in the center of the zoom, then I realized the isolines were still there heh.

I'm torn between the isolines and the brush from the first one. Both seem to have their pro's and cons. I'll have to see how things go towards the end when the tiling output is added, I'm sure he's got something up his sleeve for dealing with that problem.

Will probably be solved once layers are working again. I guess it would then simply be a case of designing each layer, saving the heightmaps without filters, saving the filter presets for each layer, removing the isolines entirely then loading the heightmaps back in. From what I can tell so far, even a standard heightmap like that, with nothing special attached to it will make use of the procedural zoom effect.

.... Ok yeah just tested it again, that does work! Needs filters added to do it though otherwise it ends up just a regular zoom, but with them, different ones on each layer for each type of feature. It should definitely speed up the process as it wouldn't have to be worrying about generating the base terrain at the same time. Excellent!


In terms of convincing geography, my advice would be to emphasise 2 features in your base map: upland areas (areas from where rivers will drain) and rivers. You'll no doubt need fewer rivers than say ME-DEM but I think that will give you a good way to orient yourself when the map starts getting crazy huge! -but more importantly (imo) it will give you believable terrain- it depends how important that is to you. If you get the big features right, GeoControl will do (most of) the rest for you.

Thats a good idea, thank you! I'm beginning to find I like to work very closely zoomed in now too. I've noticed when I can see the entire map at once, I tend not to see it as the size it is, if that makes sense? I see it as far smaller and end up adding mountains and such that are entirely the wrong scale.

Soon we could (if we have the storage/ time) be getting into the realms of terrabytes of data. The ME-DEM base map is 20K x 20K- that's really about as far as I could push the hardware I was using back then. The plan was to upscale by factors of 10 in 2 stages. So 200,000 was the immediate goal. Now that's gone by the board as a necessary plan but we'll stil probably work in powers of 10.
I'd like to get the base map to 40K *if possible*. Then that map will serve as a base camp from where we can zoom.


*faints at all the storage required* lol. Ok mine is NOWHERE near that big.. and if it gets that big I need to scrap it and start again hehehe. One of my problems is, I want people to be able to basically walk from snowy landscapes, through forests, marshes, meadows, hills, barren rockies and onto desert regions. But at a more believable distance. So far the only game I've ever seen pull that off was Daggerfall years ago. Where walking from one end to another would take weeks or something. Unfortunately it didn't look so pretty as everything was randomly generated outside of the cities. Morrowind had the detail between, but you could jump from semi-desert climate to the ice covered island in a few minutes. For mine I want people to group together in teams, preparing in advance before going out and exploring. Planning ahead, working together. It's a longshot doing that since online games are often plagued by people who don't listen hehe. But If just one group of players did that, I'd be happy.

[Jules Verne]

Morrowind was a good size. I think once you can get travel boredom, you have a believable world (though not necessarily enjoyable experience). I think the best way to guage it is to now how long it takes in real time to travel across the map. Did you ever get a look of the mmorpg Dark and Light?; they've got some great terrain in that.

The impressive part with Morrowind was the map itself, in total. Was tiny. The distance between the starting town and Balmora for example was very small really. But the way they modeled the terrain made things appear to the player to be so much larger. I remember playing it for months, often doing my own little thing with the feeling I was miles from any towns, when if I'd jumped over a few mountains. I'd find myself only next door to any town I'd need. That game really impressed me.

Unfortunately I don't have that ability. As one of the main parts of mine is the inclusion of levitation magic and flying machines (my username here kinda gives away the general style of things btw hehe) So all the familiar methods of making terrains appear larger to a player fail in this case. Which is why I've had to go with a much larger gameworld, otherwise one entire area of the game would become boring if someone could explore the entire world in a few minutes by flying around. At its current size, its going on for around the dimensions of the UK, so even with the ability to fly to any location they wish. It'll still be a long time before one person has explored everything.

I've not checked out Dark and Light. I've tried recently to avoid looking at too many upcoming games of the same genre, so I don't end up knowingly copying anything by accident.


Hehe- yep, that's where programmers come in very handy!! If you can write procedural solutions/ scripts to placing objects in your world. then you're potentially onto a real winner. Obviosuly you'd want to place any objects central to the game/plot by hand.
Yes, we're going to add everything we humanly can- but we don't have any deadlines which is just as well for an absurdly ambitious plan

Yeah. I do have a really simple way of adding lots of misc objects, randomly and sitting correctly on the terrain. Alas it ultimately creates a single object, which would cause havoc with the LOD not to mention the textures. So still working on a simplified solution to that problem hehe. Maybe all the tree's have been cut down and all the rocks picked up? lol.

If I can at some point afford a license for speed tree, that might save a lot of work. Failing that I'm sure I'll find something. Part of me thinks it will help, doing a great deal randomly. The artist in me though thinks, since many objects will be duplicates anyway, why not just create lots of versions and make things more fine tuned. Tree's built to fit their location rather than added on top and looking the same. More Fangorn style tree's than Tamriel tree's so to speak. :) I've always wanted to build a forest like that, huge giant oaks, roots all over the place, really give the location some character, you know? :)


Like I say we hope that the data (dems. 3D models, maps, etc) will stick around and be (re)used by people for all kinds of things: games, renders, online maps, etc. M-E is such a perennial favourite!
We have a developer who is working on a procedural texture generator. (Incidentally he's also building a huge world like yourself). Basically this says what type of enviroments are where based on temperature and rainfall (and we hope latitude). It spits out a 'base' map which functions rather like the base dem. Then it upsamples and stitches photos together. You could use colourmap/texture output by GeoControl or any terrain app, it's just a case of programming the compositor to know which input pixel colour corresponds to which habitat. It then goes off and finds suitable photos and stitches them together....that's the theory anway!
I'm starting to plug things here- so feel free to continue the convo along these lines on the two boards linked. Why not post your progress to the community here as these guys?:http://terrain.cg-arts.org/forum/index.php?board=22.0

monks

Pretty sure they'll still be useful years from now. Though I have heard talk of 32bit heightmaps. Which I imagine will eventually replace 16bit heightmaps. Though if you keep all your original data. I can't see it being an issue... Besides, now I think about it, doesn't GC1 already output 32bit, for World Machine?

As for the models. I suppose one way to futureproof those would be to build them in insanely high detail with massive textures. Then from those generate much lower quality versions. As time goes on, release more detailed versions of them to match what people can currently use. Another benefit there is it's much easier to scale them down than scale them up later. So it'll cut down the amount of work required at a later date.

With the storage you guys are using. Building everything in the smallest of details, creating huge sized objects isn't going to be an issue, and its a whole lot easier making the master versions like that, without having to worry about polygon counts.



Phew, all done! LOL!

[monks]

The entire map in Oblivion, displayed in a google map interface. I briefly looked at how they put it together and it too uses tiles in a very similar way to how you seem to be.

That's one of the things we have planned. :) It's taking a wee while to implement though. It'll be off the main ME-DEM page.

Exactly, from testing it I get the impression it could definitely work too. Though I have noticed in the alpha that the isolines don't follow the zooms, took me forever to work out why there were familiar markings in the center of the zoom, then
I realized the isolines were still there heh.

I seem to remember Johannes mentioned that the isolines are not functioning correctly with the zooms- that ws a couple of builds ago tho.

I'm torn between the isolines and the brush from the first one. Both seem to have their pro's and cons. I'll have to see how things go towards the end when the tiling output is added, I'm sure he's got something up his sleeve for dealing with that problem.

Hmm, yes. I'm definitely keeping version 1.0 squirelled away in a safe a place. For starters the Brush has real time feedback. Occasionally, that's the overriding requirement.

From what I can tell so far, even a standard heightmap like that, with nothing special attached to it will make use of the procedural zoom effect.

Sounds good, yes the zoom is an implementation (I think) of the 'scale and add' approach. So I'm told. Whatever it is, it's pretty awesome!




In terms of convincing geography, my advice would be to emphasise 2 features in your
base map: upland areas (areas from where rivers will drain) and rivers.

Thats a good idea, thank you! I'm beginning to find I like to work very closely zoomed in now too. I've noticed when I can see the entire map at once, I tend not to see it as the size it is, if that makes sense? I see it as far smaller and end up adding mountains and such that are entirely the wrong scale.

Well I'm sure you've done this before- generally speaking. ;) Another thing I found to pretty cool was when I was trying to get the effect that rivers were flowing from the mountains naturally- hooking uo the isoline rivers with the erosion. I did this and it worked pretty well:

generate erosion.
mark isolines so that the rivers connect up with the main erosion channels.
go back to uneroded terrain.
erode with isolines in place.



I want people to be able to basically walk from snowy landscapes, through forests, marshes, meadows, hills, barren rockies and onto desert regions. But at a more believable distance. So far the only game I've ever seen pull that off was Daggerfall years ago. Where walking from one end to another would take weeks or something. Unfortunately it didn't look so pretty as everything was randomly generated outside of the cities. Morrowind had the detail between, but you could jump from semi-desert climate to the ice covered island in a few minutes. For mine I want people to group together in teams, preparing in advance before going out and exploring. Planning ahead, working together. It's a longshot doing that since online games are often plagued by people who don't listen hehe. But If just one group of players did that, I'd be happy.

I hear you! Seriously, check out the Dark and Light mmorpg. The terrain and enviro are pretty fantastic. There's an explorer class, and I imagine it's one of the more enjoyable things to do- making/selling maps, etc.


The impressive part with Morrowind was the map itself, in total. Was tiny. The
distance between the starting town and Balmora for example was very small really. But
the way they modeled the terrain made things appear to the player to be so much larger. I remember playing it for months, often doing my own little thing with the feeling I was miles from any towns, when if I'd jumped over a few mountains. I'd find myself only next door to any town I'd need. That game really impressed me.

I was mad for Morrowind, I have to say- for many weeks lol The non-linear, sandbox
really suits me as I love to get a glimpse of the whole, rather than being tramelled too much by the narrative. Go off and explore and then hook back up with the main plot- cool. Killing Dagoth Ur was so easy for me by the time I came to do it. I was something like level 65?? I ran through the citadel with near impunity! The map in Oblivion seems smaller for some reason- maybe it is. I know that modders have had to take a hit regards terrain size with it. Oblivion doesn't (or didn't initially) support as large a terrain as Morrowind.
The very best thing about Morrowind for me was the climax to the Class 'narrative' -or
whatever you woul call it- I suppose the equivalent of becoming a lord in D&D. Gettting to live in a mage's fairytale castle was the icing on the cake.


The artist in me though thinks, since many objects will be duplicates anyway, why not just create lots of versions and make things more fine tuned. Tree's built to fit their location rather than added on top and looking the same. More Fangorn style tree's than Tamriel tree's so to speak. I've always wanted to build a forest like that, huge giant oaks, roots all over the place, really give the location some character, you know?

Exactly- enviroments are what I love. Trees and forests have a special place in my heart (even if I wasn't a Tolkien fan). I was looking at the forests in Conan (the upcoming title), they seeme do be pretty Fangorny. I'd like to specialise in trees created in ZBrush but I just don't have the time...:p


Like I say we hope that the data (dems. 3D models, maps, etc) will stick around and be (re)used by people for all kinds of things: games, renders, online maps, etc. M-E
is such a perennial favourite!

Pretty sure they'll still be useful years from now. Though I have heard talk of 32bit
heightmaps. Which I imagine will eventually replace 16bit heightmaps. Though if you
keep all your original data. I can't see it being an issue... Besides, now I think about it, doesn't GC1 already output 32bit, for World Machine? [/quote]

I'm not sure- I doubt it. At first glance 32 bit is totally unnecessary in some ways- for typical terrain modelling, but for renderers that support entire procedural planets such as TG2 one might argue a case for them. It's inevitable of course but I (and other too) think we need more communication between the terrain modellers and renderers. At the moment we have the renderers driving the standards. Hopefully we can hammer these things out ot help them along on the Terrain Summit.

As for the models. I suppose one way to future proof those would be to build them in insanely high detail with massive textures. Then from those generate much lower
quality versions. As time goes on, release more detailed versions of them to match what people can currently use. Another benefit there is it's much easier to scale them down than scale them up later. So it'll cut down the amount of work required at a later date.

Yeah, that's usually the way but I think we'll probably go the other way. Build them as basic. It's shard tosay because we don't have a definite single visualisation for them + the nature of ME-DEM is really designed to encourage an openness: modellers from all over chipping in and contributing and of course downloading and using! Maybe we'll get new software that can increase the details procedurally- I'm sure that'll happen, but not I know not when! :)

monks