http://www.kk3d.de/
Terragen 2 Preview
For those of you who want to see an animated sequence of an entire planet from far orbit to the surface check this site out… plenty of eye candy… i was talking to one of the planetside staff about entire planet render capabilities of earth-like worlds… got my wish sooner then expected… one can only try to imagine what this may look like on high quality settings with trees, rocks, fluvial erosion, cloud layers etc… this is just neat stuff…
Hmmmm… yummy!.. This is very eye-candy! it is that sort of candy that dissolves rather slowly in your eye… The type of chewing candy that forces you to see certain portions of the video again!..
I don’t mean to argue with Mat’s scenery making skills, but I think World Machine would help a lot in the realism of that planet… I wonder how WM will be aplicable to TG2’s planets… I guess this is more of a question to put on the TG forum, regarding making a planet out of heightmaps… Someone must have asked that before… I just need to look for it there…
PS - Viewing that video several times reminded me of a similar experience I had with a splashscreen, recently… 
Matt made the “world”, Karsten animated it, but Matt’s “world” was never meant to be seen from orbit. It would need lots of changes, which TGD is definitely capable of, to make it look more like a real planet.
As far as World Machine’s role, it would be extremely difficult to model an entire planet well in WM. You could do a relatively low resolution base terrain that would then be displaced by TG, but there is no facility for spherical mapping in WM (yet?), so this would cause some artifacts. Inevitably when mapping to a planet you’re going to have to deal with distortion somehow, so if WM wants to move in this direction at any point, it is something that Stephen will need to figure out.
I do agree that WM could have made some nicer landforms in isolated areas, but I would tend to approach that more in the sense of using WM to model just the general are that the camera eventually zooms down to, rather than the whole planet. You would use WM to create your “area of detail”, then model a procedural planet to fit around it, so that the seam would not be obvious when you zoomed in. Then you would have the best of both worlds - a full planet worth of terrain, but more detail and realism in the area you needed it most (your area of focus). You would probably want to use WM to form the basis for other parts of the planet though, this is true (mostly to keep consistency). It will indeed be interesting to see how we can make this stuff work and avoid undesirable distortion.
Making a planet out of hieght maps is something i will try first thing hopefully if i can beta test that puppy by years end… i have imaginary worlds going back to 1991 that need to be done… i know that matt used MOLA data of mars for the mars animation on the official planetside website so even height maps should not be an issue even layers of hightmaps… if we can import maps now in the current version then by terragen 2 we should be able to do so much more… yes the quality of this animation is OK but we are talking about the alpha version and this was rendered on low quality settings so this is simply a test of what is to come… i am sure world machine will greatly compliment terragen 2… i do not see why it can’t…
i wonder if one can use a spherical mapping plugin in photoshop… i may have afew lying around somewhere… could that solve the issue of the distortion perhaps??? and then one could add extra fine details within terragen 2 over the regular imported world machine map as we zoom in closer… does this make any sense???
I’m one of the TG2 alpha testers, so I’m speaking from experience here. 
Trust me, it will be hard to get global data working right straight out of WM. With some kind of tricky georeferencing “correction” you might be able to make it work well though. Or if Stephen implements some kind if virtual georeferencing for imaginary world mapping in WM. It would be easiest of Stephen handled it on his end I believe, although I don’t know the best way to do it.
The MOLA data is directly sampled from a real planetary body, so the distortion can easily be accounted for with known values for the planetary radius, etc. and of course the actual data is already on a sphere. WM’s data is sampled on a theoretical infinite flat plane, at least right now, so there’s not even a way to create a planet to fit it based on imaginary data - there is no inherent curvature, as far as I know. Attempting to apply known georeferencing to it would make it look very compressed toward the poles.
Anyway, I think the problem may be more difficult and complex than you realize. In real DEM data for example, with georeferencing applied to heightfield data, the northern lattitude quads have noticeably smaller tops than bottoms - the quad is far from a perfect square. Without introducing some way to account for that on the data generation side of a terrain creation system like WM, you will inevitably end up with potentially noticeable distortion. This would make it particularly difficult - even if you could get WM data to work in a planetary context - to create features in your designs that would actually look the same when mapped to your globe.
Well i think i know what i can do… i could just use mask maps edited from within photoshop imported into terragen so then i could have several layers of masks for each regional land type and climates as well so if wanted mountains i can have them rendered within that masked area only… the hieghtfields can be rendered only within those masks so then mountains, hills, seas, plains etc. can be placed in a realistic fashion… i probably could even mask the north and south poles… now if i am not mistaken we will be able to have different types of terrains displaced over layers sort of like photoshop so that masking idea should be an easy way to go about planets without the distortion troubles sense no hightfield maps will be imported… only temparary mask maps for terragen’s terrain editor… does this seem reasonable???
Maybe I’m misunderstanding you, but it seems any imported image (and that includes heightfields) would be distorted. Unless you are able to account for this distortion in your mask creation process effectively, it would be noticeable. It all comes back to a way to account for the distortion. Using UV tools available in 3D painting apps like Bodypaint or UV Mapper Pro you could probably do it fairly easily, but the source data might still need significant manipulation to work right. If you were hand-painting or assembling it to begin with, you could tailor it to an appropriate projection/UV from the start.
I’m sorry i should have been more clear to say what i wanted to do… because i know in terragen 2 you could use masks to displace terrain, clouds, surfaces, vegitation and so on… i just wanted to use that masking over an entire planet… so that is the idea… does that make sense now… sorry 
Yes, I understood that, but the masks would still be subject to the same distortion due to spherical mapping.
True, 100% spherical mapping to the planet would be almost impossible to do within the framework of WM. There’s no way to wrap a flat piece of paper around a sphere without crinkling the paper.
But you could probably acceptably map most of the globe – just the poles introduce significant problems. This isn’t enough for the kind of whole-solar-system creation that that animation shows… but its plenty good enough for any kind of on-planet shots.
Just found this thread by coincidence.
Correct, Matt created the world, but it was never designed for viewing from outer space. I just took the file and did the animation part. So don’t expect this as a standard or whatever, I would prefer the description “feasibility study”.
In general, I never tried that, but I think it would be possible to import many WM terrains into a TG2 generated procedural planetary structure, imagine it like a basketball where you stick on little squares… or a patch carpet… you know what I mean. No idea if it would really look seamless, but there are some edge smoothing algorithms.
(...) I just took the file and did the animation part. (...)In general, I never tried that, but I think it would be possible to import many WM terrains into a TG2 generated procedural planetary structure, imagine it like a basketball where you stick on little squares… or a patch carpet… you know what I mean.
- Karsten
You’re the one behind that animation? Nice job!..
I thought that was something more offitial as it had the name of Matt in it… But I keep forgetting that something looking good, is no longer a simptom of “offitial”, these days… I think I was mislead by the credits appearing all at the same time, instead of “Made by X based on work by Y and Z”…
Anyway, back to the wrapping problem in WM… I think there might be some way to fake the wrapping of tiles arround the sphere, WM-speaking… The “basket ball” goes ok up/down to the artic/antartic circles (or to the tropics, if you’re a perfectionist ;)). Then you can blend this with other maps for the polar regions… (as it has been said already)
This would require some trouble on Stephen’s part, to automate inside WM… (to do the blending via masks inseide WM in a way that the user would understand what is “polar region” and what isn’t, or have any ground reference to where/what they should be…) Given all this trouble, I’d say it would be best to do it “decently”.
Now what would “decently” be?
Would it be some sort of coordinate change that would transform a planar seed-based perlim noise into a periodic (repeatable) spherical noise? (transform height into radius, and x y to polar coordinates?)
Or would it be best to enable non-quadrilateral terrain files? like being able to make pentagonal and triangolar terrain files?
The first possibility might still have the problem of what happens when latitude is 90 or -90 degrees, as the perlim noise function must have continuity there… So polar coords may not be the best…
The second idea came to me from the fact that you can sort of create a sphere with pentagons and exagons, or with triangles… Though, this idea only aleviates the problem of shrinking the square tiles near the poles… The problem of maping generator devices to a sphere remains untackled.
In any case, you would still need some cartographic knowledge to decently display stuff in a 2D preview…
But keep in mind that once you are in a 60 degree latitude, you will have things shrunk by half comparing to the equator. So a 256x256 terrain patch with a resolution of 256 pixels per Km, would be 1Km long in the equator, and 1/2 Km at a 60 degree latitude… It would be a bit claustrofobic All those nice perfectly-round vulcanos and craters would go ugly-oval bye-bye… 
Oh well… Nice problem we’ve got our selves in!!
why can’t i just take the bitmap image from a saved world machine render and import into a spherical mapping program and then import it into terragen 2 later on… the spherical mapping software should solve the distortion were the poles are and so on… if this does not work i will eat my hat…
Well, I think it will work… It will transform something planar into something mappable to a sphere. I don’t know what exactly does such a program do, but in any case you should get something usable…
My concerns expressed earlier were seen from a perfectionist point of view… So I was thinking of what would be needed to have WM output the desired result directly…
Well, I’d be glad to test it in TG2 if you’d like to make a sample and reproject it. I can also test in Cinema 4D.
well i suppose i can render a terrain map in world machine, export it as a bitmap, import into a spherical projection program then save it again as a bitmap… does that sound good???
Sure. Let me know when and where it’s available.
i will have to e-mail it to you… is that OK???
BTW, if you can tell us what program you use to do the spherical projection…