Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[mikemayday]

Hello!
Not sure if this is the right place to write about these, sorry if not!

I've recently discovered Anteworld and I must say I am blown away by the terrain generator. There are quite a lot of graphical engines available, with enough money one can hire fantastic concept and 3d artists, I don't trust anyone other than myself to come up with good RPG gameplay anymore, but the planet-scale full-detail world generation? That's a first.
I therefore had some questions and notes of appreciation about the generator.

1.From a techincal point of view, Anteworld is looking fantastic. The level of detail, the mind-numbing scale, the atmoshpere simulation- that stuff is fantastic. Are you also planning to allocate resources to the artistic side of the world generation? Some artist to work on the textures, colours etc. The landscape often looks pretty artificial at this point- more work on the flora and geology will no doubt alleviate that, but I'm sure some artistic direction would of help. On the other hand, maybe you're not interested in allocation resources to the artistic side of what is essentially a tech demo for Outerra.

2.Biomes.
The current "european" biome is already looking up. Are you planning on perfecting it before moving on to more biomes (different geological structure, different flora) or are you going to develop them simultaneously? Personally I'd love to see the first biome perfected before work on the others begins.

3.Flora
The placeholder billboard is a work of art, I must say- had me fooled for a surprisingly long time, even up close. And at a distance? I never thought a billboard-based forest can look so realistic- the way the forests look from a distance now is already perfect IMHO!
I assume adding a varied flora will do lots to help build a sence of scale and immersion. The models angrypig posted here http://forum.outerra.com/index.php?topic=348.0 look pretty worrying though. I know a guy who made a fantastic flora mod for Mount and Blade- he mentioned he was planning to release his models in GPL someday, I think they might be worth looking into? Check these out:
http://wypierpapier.blogspot.com/2011/12/polished-landscapes.html
He's also responsible for the landscapes in Shogun 2
http://wypierpapier.blogspot.com/2012/04/shogun2-fall-of-samurai.html
I believe if you can convince him to help, the amount of imrpovement to the visual side of Anteoworld would be tremendous.

4.Geology
When looking at a single, small fragment of rocky, mountaneous terrain, a lot of what is already in place looks really well. It's often way too smooth, but it does look like a real place. On a larger scale however, everything is extremely uniform and bland. I would die happy if you guys worked on simulating the Tatras (do you share my love of our beautiful mountains?). There's lots of stuff going on here to be inspired with when making a world-simulation.
http://upload.wikimedia.org/wikipedia/commons/8/8d/Panorama-Morskiego-Oka.jpg
One of the most sorely missed elemetns in Outerran landscape IMHO are the sharp borders between gentle soil slopes and sharp, almost vertical LARGE rock formations- as seen on the left side of the photo above.
Out of all the things to place, the collovium is what worriees me the most- were you planning on implementing it in the simulation? Since it's essentially lots and lots of smaller rocks and boulders, I assume it might take a serious hit on the framerate- do you have any ideas on making it work? River bedloads and mountain lake bottoms pose the same problem.

The smaller rock formations you've got look like a great start, but it's obvious they are procedurally generated- no sharp edges, no crevices. Are you considering procedurally-placed but hand made models for rock formations?


All in all I see lots of room for easy improvement and lots of room for very difficult work. I'm no coder sadly, so I could only help with 3d modelling if you wish, but I'm rooting for the project like hell- this is something I was dreaming of helping make one day (there are hundreds of mountain photos for inspiration on my HDD . I'm pretty sure that more flora, lakes and rivers and some collovium would already help a lot to make the places much more realistic- I can only dream of how amazing things will look after more thorough geological simulation. If you're interested in some research on that matter- please let me know, I'd love to help.

Best regards and good luck,
Mike Mayday

[cameni]

Are you also planning to allocate resources to the artistic side of the world generation? Some artist to work on the textures, colours etc. The landscape often looks pretty artificial at this point- more work on the flora and geology will no doubt alleviate that, but I'm sure some artistic direction would of help. On the other hand, maybe you're not interested in allocation resources to the artistic side of what is essentially a tech demo for Outerra.

We aren't allocating resources for it because the whole system is not in place yet. Vegetation needs a generator capable to produce all LODs for OT so that it doesn't suffocate under the load, if it was rendered naively. But with terrain tools and procedural biome vegetation support it will start looking much more interesting. There will be some artistic input needed - definitions for the vegetation and its distribution rules etc, but right now the system is very simple.

The current "european" biome is already looking up. Are you planning on perfecting it before moving on to more biomes (different geological structure, different flora) or are you going to develop them simultaneously? Personally I'd love to see the first biome perfected before work on the others begins.

The whole system is largely procedural, so a work on one biome will enable the same enhancements in other biomes. Of course, that's for the baseline stuff - some assets that make up the details will be sooner available for one biome. So it's kind of both

The placeholder billboard is a work of art, I must say- had me fooled for a surprisingly long time, even up close. And at a distance? I never thought a billboard-based forest can look so realistic- the way the forests look from a distance now is already perfect IMHO!
I assume adding a varied flora will do lots to help build a sence of scale and immersion. The models angrypig posted here http://forum.outerra.com/index.php?topic=348.0 look pretty worrying though. I know a guy who made a fantastic flora mod for Mount and Blade- he mentioned he was planning to release his models in GPL someday, I think they might be worth looking into? Check these out:
...
I believe if you can convince him to help, the amount of imrpovement to the visual side of Anteoworld would be tremendous.

The problem is that trees must be able to transition to billboards as seamlessly as possible, and they must lose detail as soon as it's possible because in the open world and with the visibility like in OT there's a huge amount of them, and the performance is obviously a huge concern.

That's why we are working on a tree generator: it can produce all the necessary LOD levels and the most fitting billboard, so that the transition is minimized and switching least visible; it also generates tree data in the optimal format for a specialized renderer. While it would be possible to make these hand-modeled, it would be quite tiresome and hard to even explain to an artist.

When looking at a single, small fragment of rocky, mountaneous terrain, a lot of what is already in place looks really well. It's often way too smooth, but it does look like a real place. On a larger scale however, everything is extremely uniform and bland. I would die happy if you guys worked on simulating the Tatras
...
One of the most sorely missed elemetns in Outerran landscape IMHO are the sharp borders between gentle soil slopes and sharp, almost vertical LARGE rock formations- as seen on the left side of the photo above.

This is possible, you can see sharp transitions in some places but in smaller scale, where it was generated by fractals. However, 90m world data grid does not capture the the sharp transitions, and it's a huge difference when compared to 30m data, especially in mountains. But even with 90m data there are tricks that could be used to enhance the terrain spectrum; right now the extra fractal detail between 100 and 30m is quite weak, not bringing much of a variety at that part of the spectrum. This was especially visible when I recently tried Mars, using 500m data - the lack of artificially generated details above 30m was even more visible there. But that's simply because we need to tune the fractals for that part better.

Out of all the things to place, the collovium is what worriees me the most- were you planning on implementing it in the simulation? Since it's essentially lots and lots of smaller rocks and boulders, I assume it might take a serious hit on the framerate- do you have any ideas on making it work? River bedloads and mountain lake bottoms pose the same problem.

The smaller rock formations you've got look like a great start, but it's obvious they are procedurally generated- no sharp edges, no crevices. Are you considering procedurally-placed but hand made models for rock formations?

Those rocks are a part of the terrain and thus do not cause extra performance loss. There's a lot that can be done with the procedural terrain generator, but it also places some constraints - mainly because it should not get too complicated, as its used everywhere. But there's also a vector overlay system that has better capabilities and a bigger freedom, since it's used only on selected parts. It's used for roads, craters, will be used for rivers, pasture and field placement etc. I think it can also be used for the colluvium, as it's able to alter shapes and materials of the underlying terrain using pattern lookups, so it should be also possible to create that - but it will need some experimenting there.
But yes, visiting Tatras is .. depressive, from the point of view that how much work there is

[2eyed]

Cameni, what do you think of making a certain region (Mt Rainier or Grand Canyon) in higher dem resolution, maybe 30m or even 10m? Could OT handle the additional load and how would the borders to normal (90m) terrain look like?

[cameni]

It's possible but the transitions to the surrounding coarser terrain needs to be solved first. Technically this already works, as the land data are of higher precision than the oceans depths, but the approach used there isn't suitable for land transitions as it loses some of the fractal generated detail in the coarser regions.

[mikemayday]

Hey there, thanks for your responses!

The problem is that trees must be able to transition to billboards as seamlessly as possible, and they must lose detail as soon as it's possible because in the open world and with the visibility like in OT there's a huge amount of them, and the performance is obviously a huge concern.

No discussion there.

That's why we are working on a tree generator: it can produce all the necessary LOD levels and the most fitting billboard, so that the transition is minimized and switching least visible; it also generates tree data in the optimal format for a specialized renderer. While it would be possible to make these hand-modeled, it would be quite tiresome and hard to even explain to an artist.

So wait, are you saying, you're shooting for something that produces the basic model, all the LODs and the billboard on the fly?? Or just a tool that produces all these to later be implemented in the standard way?
I mean, if you can pull it off and it looks good- that would be amazing, it's just that I have very little faith in a good-looking, 100% procedurally generated low poly tree model... and was wondering about some backup options.

This is possible, you can see sharp transitions in some places but in smaller scale, where it was generated by fractals. However, 90m world data grid does not capture the the sharp transitions, and it's a huge difference when compared to 30m data, especially in mountains. But even with 90m data there are tricks that could be used to enhance the terrain spectrum; right now the extra fractal detail between 100 and 30m is quite weak, not bringing much of a variety at that part of the spectrum. This was especially visible when I recently tried Mars, using 500m data - the lack of artificially generated details above 30m was even more visible there. But that's simply because we need to tune the fractals for that part better.

Alright, I'm glad the problem is acknowledged

But there's also a vector overlay system that has better capabilities and a bigger freedom, since it's used only on selected parts. It's used for roads, craters, will be used for rivers, pasture and field placement etc. I think it can also be used for the colluvium, as it's able to alter shapes and materials of the underlying terrain using pattern lookups, so it should be also possible to create that - but it will need some experimenting there.

Just making sure, we're clear, I meant something like this- http://mw2.google.com/mw-panoramio/photos/medium/66936386.jpg
So that looks like it would require a lot more vertices to render from up close than the rocks you've got in place right now.

But yes, visiting Tatras is .. depressive, from the point of view that how much work there is

Ah, I hope you can switch from depressive to inspiring

Cameni, what do you think of making a certain region (Mt Rainier or Grand Canyon) in higher dem resolution, maybe 30m or even 10m?

Right now, when the fractal generator is rather rudimental, I'd also LOVE to see this. And even later on, there will always be a use for this.
However, if you're also shooting for a world generator, the more realistic and detailed terrain you can create with as minimal data as possible- the more powerful the tool and the less work for the end user.


I wanted to talk about something else, but first let me reiterate- I am neither a geologist, nor a botanist, nor a programmer. Barely a graphic artist and not a top-class one. However, generation of virtual worlds is a topic I am extremely passionate about and I've spent several years gathering various reference materials including those on how landscape is formed (geology, flora, climate, etc.). I know some programmers who were interested in starting a project like this, but we're still waiting until we have the time.

Now I know this isn't the kind of help one usually looks for in a project (as it borders on telling people what to do rather than helping them do it), but if you think it will be put to good use, I could organize and share my reference material regarding landscape creation (from the theoretical side, which means "what exactly should be generated and where", sadly not "how to code it").

A follow-up to my question about the biomes- the reason I asked is because I believe that the best approach to making interesting terrain is determining what elements appear in a biome and creating different procedures that generate these elements, then placing those elements procedurally- as opposed to what is currently there, which is, I assume, once procedure to handle all terrain. And naturally, the biome for the Tatras would require vastly different procedures than the Grand Canyon. UNLESS of course, you're willing to create a procedure so accurate, that it will produce realistic results based on the parameters of the geological structure-in which case I'll bow myself out, sit down and observe in amazement

[cameni]

So wait, are you saying, you're shooting for something that produces the basic model, all the LODs and the billboard on the fly?? Or just a tool that produces all these to later be implemented in the standard way?
I mean, if you can pull it off and it looks good- that would be amazing, it's just that I have very little faith in a good-looking, 100% procedurally generated low poly tree model... and was wondering about some backup options.

That's procedural in the sense of using rules and provided base elements to make the trees. Procedural rules allow us to produce detail levels automatically and in the format we need, while also allowing to vary the output by randomizing around the pivot values.
Technically it's possible to create the models manually, just as you'd make any other model with LODs, plus a billboard at the end of the chain. However, it can be expected that this will be a lot slower and not suitable for the forests in OT.

A logical thing to do is to combine the two approaches. Inside the deep forests, trees are getting more unified in the looks, and it really doesn't matter that much if the gazillionth tree you encounter looks any special. Standalone trees, and trees at the forest boundaries can be made custom. But still, a custom made tree is just a fixed one, and you'd need several models. That's increasingly taxing on the resources too, whereas procedurally defined trees use a constant memory.

Just making sure, we're clear, I meant something like this- http://mw2.google.com/mw-panoramio/photos/medium/66936386.jpg
So that looks like it would require a lot more vertices to render from up close than the rocks you've got in place right now.

Yea, definitely. I think I already described it elsewhere, what we want to do is to use the vector overlay generator to produce an envelope of this, so that it can be used for physics without modifications, and would also serve for the more distant depth levels. Closer to the camera it would transition to procedurally generated rocks, so that there are cavities and all.
Much like the grass, actually.

I could organize and share my reference material regarding landscape creation (from the theoretical side, which means "what exactly should be generated and where", sadly not "how to code it").

I'm sure a website with this info would be highly valuable to anyone doing terrain modelling. Most of what I wished for Outerra was inspired by the views in my home land (Orava), and how the vegetation is distributed there. Even just a collection of geotypical imagery for major regions would be of help here.

A follow-up to my question about the biomes- the reason I asked is because I believe that the best approach to making interesting terrain is determining what elements appear in a biome and creating different procedures that generate these elements, then placing those elements procedurally- as opposed to what is currently there, which is, I assume, once procedure to handle all terrain. And naturally, the biome for the Tatras would require vastly different procedures than the Grand Canyon. UNLESS of course, you're willing to create a procedure so accurate, that it will produce realistic results based on the parameters of the geological structure-in which case I'll bow myself out, sit down and observe in amazement

Different procedures are problematic for several things. Performance (lots of branching); the need to transition between two or more sets of procedures when crossing into a different region or (worse) when having to make an aggregated terrain tile detail. Some of it can be decomposed into a multi-pass approach, some parametrized, but you always have to do a lot of compromises for the real time stuff ...

[mikemayday]

Different procedures are problematic for several things. Performance (lots of branching); the need to transition between two or more sets of procedures when crossing into a different region or (worse) when having to make an aggregated terrain tile detail. Some of it can be decomposed into a multi-pass approach, some parametrized, but you always have to do a lot of compromises for the real time stuff ...

I think it helps that the places where the placement of elements matters would be mountainous regions- and as far as I know, two different types of mountain biomes don't border on each other without an area  of generic mild hills or flatlands between them.

But all in all I just really hope you guys strive towards ending up with couloirs, arêtes, colluvial cones and more realistic rock formations- I sure ain't the right person to tell you how or IF to do it at all. I'll gladly stay and watch and let you guys know when I organize my materials!

[HiFlyer]

When I look at the potentials for Outerra I look at other similar engines with their different approaches and how they are proceeding and solving things in faith that if it can be solved in a similar engine, then it should be at least theoretically repeatable in Outerra.

So! On to rock formations!

I saw an interesting rock  experiment that was very promising in that regard. I don't think there is any reason why Outerra could not match similar terrain features........