Celestia Forums

At first glance you would think this is a cloudmap. But it's not really. It's a 3d model with a cloudmap texture. It is also using a normalmap and specmap. The earth is Celestia's default earth sphere with my BMNG July texture, normal and specmap. These screenhots show one of the effects as the light move across the planet. I'm using OGL2 and cvs 08/10/2006 with cloudmaps off. Except the last shot has cloudmaps and cloud shadows on.








cartrite

A couple of screenhots from the more detailed model. about 3 million verts.



cartrite

That is looking very nice Cartrite. It is the best thing to happen to clouds in Celestia since I cam up with the idea of using the overlay texture for creating cloud shadows. I would love to see my Summer cloud texture treated to something like this. At what equicalant level ar these models.
8k or 16k? If you would be inteerested I can post my Summer cloudmap at the 16k level for you to try. It is doing nothing but sitting in a folder on my system at the present time totaly unused.

Don. Edwards

Hi Don, Thank You,
I just came up with this idea today so this is all preliminary. What I did was bump the 8k cloud map from the old bluemarble site with the gimp and used it to create the model with blender and Toti's z-fold script. Then I used the same map to make a normalmap with the gimp. Kind of crude but effective. I guess the same can be done with your maps.

The model is using 4k maps in the screenshots.

If the overlay texture is a bump map and matches the Summer cloud texture then half the job is done. The model can only be about 131k faces in blender and that becomes 262k triangles when it's converted to cmod. That equates to about 6 to 10 fps on my card which is an old one. If I really took my time I could remove faces that were not being used and add to the detail where it was needed.

What I started doing was dividing the model into parts and assigning each a texture. That way I can use a 16k or larger as a basetexture on the model. It takes time to do that though because each tile needs to be overlapped to minimize the seams.

But the problem is that this may be all temporary because of the paragraph below. If the code gets fixed this may not work anymore.

The new cvs code that enables normalmaps and per pixel lighting has an annomally that when a texture is used that has an alpha channel (with nothing in the alpha channel) the texture is drawn with about 50 % opacity and the model becomes transparent and whats in the background shows through. In this case the earth. I guess it's a bug but it's a good one. I hope it don't go away. It could be benificial.

A good example for what I mean about the model being transparent are the 2nd and 3rd thumbs in my first post. You can see the coastline. If cloudmaps were enabled you would only see the cloud.

cartrite

Nice trick indeed!
Will you make the model available?

Nice way to work around the lack of normal maps for cloud textures!

cartrite wrote:But the problem is that this may be all temporary because of the paragraph below. If the code gets fixed this may not work anymore.

The new cvs code that enables normalmaps and per pixel lighting has an annomally that when a texture is used that has an alpha channel (with nothing in the alpha channel) the texture is drawn with about 50 % opacity and the model becomes transparent and whats in the background shows through. In this case the earth. I guess it's a bug but it's a good one. I hope it don't go away. It could be benificial.

It is a bug, and I think that it's already been fixed in the CVS code. However, you can achieve the same effect by setting the correct alpha values in the texture map, just like you do for a regular cloud map. In fact, this should give a better result than setting a single opacity for the entire cloud map.

A good example for what I mean about the model being transparent are the 2nd and 3rd thumbs in my first post. You can see the coastline. If cloudmaps were enabled you would only see the cloud.

You can achieve this partial transparency with regular cloud maps, too. But most of the cloud maps out there have large regions of clouds that are almost completely opaque (which is more realistic, though your more transparent clouds are interesting.)

--Chris

Chris wrote:

You can achieve this partial transparency with regular cloud maps, too. But most of the cloud maps out there have large regions of clouds that are almost completely opaque (which is more realistic, though your more transparent clouds are interesting.)

I just seen that this works as I can now lower the opacity in the other rendering modes and make them semi transparent. The clouds are opaque when opacity is set to 1 in the model. All except OGL2.

I think that what I'm going to try is having a low resoloution model with a semi transparent cloud layer thats normalmapped and search for some data in Ir for thunderstorms or huricanes and create a 3d model of a huricane and a couple different thunderstorms to embed in the lores mesh. These would be opaque.

ElChristou wrote:

Will you make the model available?

If I get a finished model,yes. I just started with this and I'm trying a lot of different things.
so it will be a little while. The model is easy to make tho. If you use blender, Toti's zfold script does most of the work. I'm sure there are other methods for other modeling software too.

cartrite

Cartrite,

I can't help wondering what the performance of this model will be like on low-spec machines (as that's the situation I'm in) -- You said it was 3 million polygons I think!!

Any idea?

chris wrote:Nice way to work around the lack of normal maps for cloud textures!

...

--Chris

Sorry for being a bit less enthusiastic here...

For my taste Cartrite's clouds don't resemble REAL clouds much. My old 2k clouds that were used by many over the years were given their "fluffy" appearance precisely by means of (static) bump mapping...So that technology is not new in case of clouds. A normalmap for clouds would admittedly make the bumpyness dependent on the sun's position. Fine.

From looking at many satellite photos of clouds it seems to me that from space they appear very much like a fluffy /2d/ layer rather than like a fully developped 3d structure (that would require a model). Of course that changes if the observer is getting very close, diving into the clouds, actually.

Bye Fridger

Chuft-Captain wrote:

I can't help wondering what the performance of this model will be like on low-spec machines

I only built this computer from the remains of my former computer that was fried by overwork not too long ago. That was a used amd k7 900mhz with 512 mb ram that was donated to me from a trash bin. With it, I was able to create a 64k blue marble texture and create models of similar size to name a few and although the performance was nothing to write home about, they still ran. The system I have now is not that much better, with performance that is. I can work on larger files and eliminated some steps that were necesary with my old computer but I still use the old card and my performance in Celestia has not changed that much.

My guess is if your machine can look at a 4k texture, the model from my first post would work fine. The model with 3 million verts also brings my machine to its knees and was only created to help me get a feeling for a benchmark. My plan is to use the smaller model in pieces so I can increase texture size. If I cut into 8 pieces square then a 16k texture set could be used on machines that can handle 4k textures. An 8k if a machine can only handle 2k textures and 4k if it can only handle a 1k texture.

cartrite

Fridger wrote:

chris wrote:
Nice way to work around the lack of normal maps for cloud textures!

...

--Chris


Sorry for being a bit less enthusiastic here...

For my taste Cartrite's clouds don't resemble REAL clouds much. My old 2k clouds that were used by many over the years were given their "fluffy" appearance precisely by means of (static) bump mapping...So that technology is not new in case of clouds. A normalmap for clouds would admittedly make the bumpyness dependent on the sun's position. Fine.

The model does not have to have any real 3d structure for the majority of the cloud map.
It's just a platform so I can use a normalmap to draw sun dependent shadows on the cloud. Thats what I'm really trying to do. Some links to Astronaut photography from space that do show 3d shadow structure.

http://eol.jsc.nasa.gov/scripts/sseop/Q ... 13-E-43846
http://eol.jsc.nasa.gov/scripts/sseop/Q ... 13-E-43849
http://eol.jsc.nasa.gov/scripts/sseop/Q ... 13-E-43962
http://eol.jsc.nasa.gov/scripts/sseop/Q ... 13-E-44109
http://eol.jsc.nasa.gov/scripts/sseop/Q ... 13-E-44281
http://eol.jsc.nasa.gov/scripts/sseop/Q ... 13-E-44312

Those screenshots show what I would like to do by creating a model that shows a little detail so that shadows could be drawn and high resoloution meshes of thunderstorms embedded in the model for 3d effect.
cartrite

cartrite wrote:Fridger wrote:

chris wrote:
Nice way to work around the lack of normal maps for cloud textures!

...

--Chris


Sorry for being a bit less enthusiastic here...

For my taste Cartrite's clouds don't resemble REAL clouds much. My old 2k clouds that were used by many over the years were given their "fluffy" appearance precisely by means of (static) bump mapping...So that technology is not new in case of clouds. A normalmap for clouds would admittedly make the bumpyness dependent on the sun's position. Fine.

The model does not have to have any real 3d structure for the majority of the cloud map.
It's just a platform so I can use a normalmap to draw sun dependent shadows on the cloud. Thats what I'm really trying to do. Some links to Astronaut photography from space that do show 3d shadow structure.

...

Those screenshots show what I would like to do by creating a model that shows a little detail so that shadows could be drawn and high resoloution meshes of thunderstorms embedded in the model for 3d effect.
cartrite

Oh, I see...You are using of course Chris' new code where normal maps are applicable to models. But so one might wrap the cloud texture around a very simple spherical shell model? That's presumably enough for many purposes?

Bye Fridger

Fridger wrote:

Oh, I see...You are using of course Chris' new code where normal maps are applicable to models. But so one might wrap the cloud texture around a very simple spherical shell model? That's presumably enough for many purposes?

Correct, Thats the only way I can figure to use a normalmap with a cloudmap. Also notice that it's using the new atmosphere code. It's cvs 08/10/2006. Anyhow, most of the model would have a wave structure where the low points = noclouds and high points = clouds. Not much detail, just enough to make the transitions smooth. The detail will come from the textures and normalmaps. I figure I can use a 16k texture as I did with mars. The eight section model approach can have a 1k texture on one model and a 4k on the others. Still some can have 2k textures.

The thunderstroms would be modeled so that they have a top and bottom and will be placed into squaw lines. So they could be viewed from any direction with lots of detail.. That can be a little tricky I think. Never done that before. Hope it can be done.

cartrite

A first attempt at modeling a hurricane. The data was obtained from the NOAA site and the hurricane is Katrina. I think I may have scaled it down too much. I still got a lot of work on this but it's getting there.






cartrite

You are getting close but I do think you are right, the scale is a bit small. My offer of the 16k texure still stands. It of course already has the storms. Just let me now and I will post it for you. Just tell me what size you want to work with.

Hey Don,

Thank You, I'll give your map a try. My map is kind of raw. What I will need to do is make height maps to create the model. The 16k would probally have the most detail for the heightmaps.

I goofed when I scaled that model. It gets kind of tricky when looking at thousands of little boxes and trying to decide which ones should be used and which ones should be deleted.

Another problem I was having is the data images. I downloaded many images of hurricanes from NASA/NOAA and every one has outlines around the land and some even the state boundries. A chore to deal with.



cartrite

As soon as I am back on my main machine and done fooling around in Vista I will upload both the tranparent 16k and the black and white base image so you cam use it to create the bumpmap. I wil PM you with the URL.

Don.

Meet Isabel a 4k map + 4k normalmap on a simple sphere. The specmap is a simple gray image, all 0x40's. The radius of the model is 6385.140 to match the cloud height of 7.. The model,cloudmap and cloudshadows are on. I had to darken the cloudmap and darken the diffuse color in the model because of an additive effect that turns the clouds too bright when everything is on. I also changed MieAsymmetry to -0.85.
Basicly a workaround for the semitransparent clouds on the model in OGL2.

cartrite

Don. Edwards wrote:Its looking good. But now its a bit on the big size now.

A bit? lol...