Has anything good been done with Atlas yet?

#1 The previous version of Atlas which didn't use the current ClipMap texture setup had alot better detail. In the end, it should be similiar, but somewhere along the way the texture isn't being used at its best quality it would seem.

I'm working on figuring some of it out with our Mappers, as they like to hate on the loss of Quality which is very evident.

#2 Most of the screenshots here (almost all) were taken without detail maps, and thus look blurry. Atlas with detail maps looks far better than legacy terrain, runs faster, and can be much larger. Atlas has had detail texturing though since MS4.2, there just aren't many screenshots of it.

#3 I would agree that detail map added back alot of the missing detail, but in the end Atlas looked far better than Atlas2 in most cases, at least for us.

Its getting there, as the proper working detail map texturing is a good example.. Although there are still issues with lighting/detail texturing and how it affects/lightens up the terrain..

#4 As well there is a resource which allows you to add a second detail texture, this makes another large difference.

My main prob with Atlas 2 currently is that it seems to be highly ignorant to any kind of light information or shadowing ... but perhaps I'm just too stupid to set it up correctly right now.

#5 Ok is it possible to have more than one deatail texture? or is it just one that tile over the whole surface?
for instance can the detail texture for grass areas be different than the detail texture for dirt areas?


Second is can there be more than one level of detail textures? for instance at one distance you have no detail you get closer and you have the regure deatil texture. is it possible to have an eve finier level map for the areas very close to the player?

#6 I've taken stabs at both of what your talking about James. I really want to get it to use a different detail for each area, but I've found that to be very difficult to implement. I've run low on time again but I did give it a solid two days and could not get it working.

Using two detail textures based on distance though isn't too hard. I did get it working, without much modification on the two detail code I posted last week. The problem is that I couldn't get it working right for pixel shader 1.1, without making another shader at least. It worked fine for ps2.0. Basically just in the area where you fill the renderdetail list remove the check of distance, so it renders details to the entire terrain. And then in your pixel shader code you render the first texture lerping it as normal, and you render the second texture for everythign beyond the EndFadeDistance. For 1.1 you'd have to modify that to use two passes. That's pretty simple to do though. I used it to put some a rocky shrubby detail in the distance and then had a finer detail in close.

If anyone manages to get the other style working though I'd be willing to fork over some cash for that code. It does not look to be very easy to do, and would likely also be difficult to manage on 1.1 cards.

#7 The only way (I know of) to get different detail textures for different land types is to implement separate Atlas terrain instances.

Now, I'm not as familiar with the Legacy terrain, but we've been very happy with Atlas so far. For us, Atlas2 has looked better than Atlas did. The original Atlas had "seams" in the terrain texture where the chunks did not match up. The biggest feature for us is the size allowed by Atlas. Our terrain is current 32k x 32k (game units) and there's no performance problems. Also, understand that the results that Atlas gives depend greatly on what is put into it. I don't know what size of textures people are using, but I have seen it recommended that the texture image be at least 4 times as large as the terrain mesh. I'm not sure if everyone is following this. I'll admit that before detail textures came back, Atlas did look kind of blah.

The biggest drawback to using Atlas right now is the lack of an editor. If you absolutely must edit your terrain in-game then you should probably stay away from Atlas right now. I'll post some screenshots of our terrain once I get back home, I think it looks pretty good.

#8 You could do different land types for each area without having multiple atlas instances, but it would be a lot of work. It would require heavy source modification.

#9 What ever I do needs to be farly easy as I'm not really a coder.

@mark
1: Could you post some information about your art pipeline as far as atlas2 is concerned?
2: So how many hightmaps is that?
3: how big in pixels are they each and how many was the original

I don't think I need 32kx32k of terrain though I don't see us ever needing more than maybe 5x5k. but nice to know it is posisible.

#10 Our art pipeline for Atlas is basically L3DT. It's tool that can randomly generate a map or take an existing map, and create a texture for it.

Our 32k terrain was generated from one heightmap. It was an 8192 square image file, originally. We would have liked to have it 2x or 4x as large as that, but 8192 was the largest map that L3DT could import. We experimented with cutting the map up and importing each separately but because of errors/differences in the Atlas import process, the edges of the maps would never perfectly line up.

The original heightmap was a PNG, but the Atlas import requires it (for conversion) to be a 16bit RAW file. Fortunately, L3DT supported this export. We generated our texture at a 4x HF/TX ratio. Yes, that means we have a 32,768 square texture file. An image file of this size needs to be split into "tiles", which L3DT also supports, so it was exported in tiles 1024 px on a side; a 32x32 grid of tiles.

Once we had our RAW, and our "tiled" texture, we just called the appropriate script functions to convert them into an Atlas2 terrain:

atlasOldGenerateChunkFileFromRaw16("Visions/data/terrains/Jerusalem.raw", 
                                       8192, 4, 1500/65536, 
                                       "Visions/data/terrains/Jerusalem.chu", 
                                       15, 4);
importOldAtlasCHU("Visions/data/terrains/Jerusalem.chu", "Visions/data/terrains/JerusalemGM.atlas");
atlasGenerateTextureTOCFromTiles(32, "Visions/data/terrains/TextureJPG/Jerusalem_x%dy%d", "Visions/data/terrains/JerusalemTX.atlas", 0);
atlasGenerateUniqueTerrain("Visions/data/terrains/Israel.atlas", "Visions/data/terrains/JerusalemGM.atlas", "Visions/data/terrains/JerusalemTX.atlas");

There's documentation on these functions on TDN. But if you would like the full "reasoning" behind how we arrived at our parameters, I can share that too.

#11 If you would like an example of a larger terrain that can be created using L3DT and Atlas2, I've posted downloadable example in this thread.

#12 I really didn't feel like purchasing another 3rd party application to be able to use the engine which I purchased, (ie. Atlas), so I'm stuck with legacy terrain for now. Before someone says there's a free version of L3DT, then do note it has limits on the size and texturing features.

#13 The purchased version has limits as well. That's the price you pay for having a program automatically generate a map texture for you. If you create the heightmap and texture yourself in PhotoShop or whatever you could get around a lot of those limits. You don't have to purchase L3DT to create Atlas maps, but it makes things easier for the artist-short teams like ours.

Yes, I agree its kind of disappointing that you need to use any sort of 3rd-party tool to create Atlas terrains. A terrain editor (and thus creator) was "promised" originally, back in the EA days, for Atlas but it hasn't materialized yet. So, again, if in-game editing is a "must have" for your terrain needs... then stick with GeoTerrain.

For us, however, we were always planning on using a pre-existing heightmap (real-world data), so it was a natural fit for us to use L3DT and Atlas.

#14 You're not being realistic.

Painting heightmaps and textures in Photoshop for a large terrain takes ages. If I'm going to restrict myself to a heightmap or stick to smaller terrains then the whole point with a new terrain engine is gone and I might as well use legacy.

I don't want to automatically generate textures or terrain. I want to be able to excavate/lower/smooth and paint it with blended textures just like legacy. It doesn't really matter if someone thinks pregenerated terrains are the way to go, because editors were announced since the first official milestone which was the one I (and many others) was looking at when I made my purchase.

I'm just saying that for now I'm staying with legacy and interiors. Works great! :)

#15 What are the pros/cons of L3DT? Limitations?

#16 Stefan: Sorry, but you aren't being realistic.

Do you want Max/Maya as well as Constructor free with your engine purchase as well? Those are 3rd party tools that are "needed" to make art assets for Torque, as well as pretty much every other engine as well.

#17 Ummm... woah Stephen. We all understand that you guys are working really hard. I mean, I've seen some of the timestamps on Ben Garney's posts, sometimes I wonder if the guy ever goes home.

What I am seeing out of Atlas right now is awesome. We have a terrain running 8 times larger than the maximum allowable under TGE. The performance out of it is also fantastic. At the same time, editors were promised for Atlas, originally. I understand they're still "on the table" for future development but right now we don't have them. I think what Stefan is seeing is, essentially, a loss of features going from GeoTerrain to Atlas.

I mean, the whole fact that GeoTerrain is still included in TGEA implies to me that you guys are acknowledging the limitations of Atlas and providing an alternative to those who don't want to use it, like Stefan. It's no big deal.

#18

Quote:
Stefan: Sorry, but you aren't being realistic.

Do you want Max/Maya as well as Constructor free with your engine purchase as well? Those are 3rd party tools that are "needed" to make art assets for Torque, as well as pretty much every other engine as well.

Atlas Editors were on the first milestone list, unlike 3D Studio MAX or Constructor. I think your comparasion is flawed and that you're trying to twist my post (I never said I wanted tools for free) into trolling when it contains perfectly valid statements that really had nothing to do with your reply.

#19 Atlas is for sure hard to use and understand at the moment. GG is working on that, but if you know what your doing you can make some incredible looking terrains. I was a doubter too until i learned to use it properly.

I'll try to post some video in the next few days of one we have in progress. It is 6.5KM island, has 130K x 130K texture, we use a 8KM view distance, no visible repetition, no detail map (at least not in the traditional sense), and is 11MB zipped on disk. It currently runs at about 100fps.

For comparison QuakeWars only boasts a 32K x 32K terrain texture on a 1.6KM terrain... weak. ;)

#20

Quote:
Ummm... woah Stephen. We all understand that you guys are working really hard. I mean, I've seen some of the timestamps on Ben Garney's posts, sometimes I wonder if the guy ever goes home.

I too understand Ben/Brian/Pat and whoever works on the TGEA core that they are doing as much as they can and that GG does what it can within it's financial limits.

Quote:
I mean, the whole fact that GeoTerrain is still included in TGEA implies to me that you guys are acknowledging the limitations of Atlas and providing an alternative to those who don't want to use it, like Stefan. It's no big deal.

Yes! I love using GeoTerrain and I will continue to do so until I see (if ever) Atlas being able to cover the same editing features (the brush ones, not the generation ones).

BUT the sad part is if someone has this opinion, they get confused with a troll or someone that wants to put a shadow on TGEA or GG.

Quote:
I was a doubter too until i learned to use it properly.

We have been able to make terrains in Atlas that were insanely more beautiful than GeoTerrain (Legacy). This is not the issue. The issue appears when you want to edit the terrain and make it conform to the rest of your world, ie interiors.

The current pipeline is enough for some teams (like Mark's) and I'm glad to hear that. But for us it didn't work out that way and we will continue to use legacy, and we're quite happy with it.

What I don't understand is the need to jump on my face because I said that, Stephen.