Christoph Hormann wrote:

> Simon de Vet wrote:
>
> > I am planning, however, to create some POV code to generate semi-realistic
> > clouds procedurally. In my experimentation, I have made some progress, and I
> > think it should be do-able.
> >
>
> But not with that rendering time :-)

I'm not so sure about this.

I've looked at other people's cloud techniques, and they all use 3-d to generate
3-d looking clouds. Some use media, or isosurfaces, or stacked planes, but all use
actuall objects.

When I look at the clouds around me, most are way too far away to look 3-d. 3-d
objects are usually shaded based on the angle of the incomming light (diffuse
shading) and on self shadowing. Clouds, on the other hand, have a bit of this, but
the vast majority of the shading comes from the extinction of light as it passes
through the water vapour.

I've been looking at clouds not as 3-d objects, but as 2-d objects, painted onto a
giant sphere. If treated this way, I think that most of the effects can be
generated with complex textures on a skysphere, which are very quick to render.

http://freespace.virgin.net/hugo.elias/models/m_main.htm has a good (award
winning) example of a pigment based cloud render.

In addition, pigments can generate clouds other than puffy cumulus, like cirrus,
stratus, and realistic haze.


I'm attaching a test image I made before I gave up on this technique, due to time
limitations. It's a very early image, with no pseudo-3-d effects, but it shows how
pigments can be used to create the shape of distant thunderclouds. It's all one
pigment, generated with the 'pigment control trick' and MegaPov's spline.


Simon

Post a reply to this message

Attachments:
Download 'cloudtest1.jpg' (43 KB)

Preview of image 'cloudtest1.jpg'

A total cheat, mister!

Sure it looks good, but imagemaps are not very flexible, cause they are
not very real... Okay I use them too, but I have found that the greatest
realism is achieved when objects are modelled like in real life.

Well, your clouds are in the horisont, so unless you plan an animation
of them, or a sundown, it doesn't matter if they're real or not. And as
long as you put enough real things into the picture, people will look at
mostly them, and not the clouds... In which case you can receieve praise
with a clean conscience.


- Hugo


Simon wrote:

>> I guess this shows that, on occasion, imagemaps aren't the terrible
>> solution they're made out to be.

Post a reply to this message

Simon de Vet wrote:
> 
> Christoph Hormann wrote:
> 
> >
> > But not with that rendering time :-)
> 
> I'm not so sure about this.
> 
> I've looked at other people's cloud techniques, and they all use 3-d to generate
> 3-d looking clouds. Some use media, or isosurfaces, or stacked planes, but all use
> actuall objects.
> 
> When I look at the clouds around me, most are way too far away to look 3-d. 3-d
> objects are usually shaded based on the angle of the incomming light (diffuse
> shading) and on self shadowing. Clouds, on the other hand, have a bit of this, but
> the vast majority of the shading comes from the extinction of light as it passes
> through the water vapour.
> 

Your description of the clouds lighting seems perfectly right, but i don't agree
about the distance.  Clouds like the ones in your scene are about 3-6 km high. 
In a large scale movement animation or a fast motion film you could perfectly
see the 3d structure of those clouds.  

> I've been looking at clouds not as 3-d objects, but as 2-d objects, painted onto a
> giant sphere. If treated this way, I think that most of the effects can be
> generated with complex textures on a skysphere, which are very quick to render.
> 
> http://freespace.virgin.net/hugo.elias/models/m_main.htm has a good (award
> winning) example of a pigment based cloud render.
> 

I have seen that one before and i like it too, but NTL, the lighting used there
is not very convincing for roundish clouds, because it only makes clouds lighter
according to their general distance to the sun, while each cloud should be
somewhat lighter on it's sun side and dark on the opposite side.  

> In addition, pigments can generate clouds other than puffy cumulus, like cirrus,
> stratus, and realistic haze.

I agree, especially for high cirrus clouds the plain texture approach seems most
efficient.  

> 
> I'm attaching a test image I made before I gave up on this technique, due to time
> limitations. It's a very early image, with no pseudo-3-d effects, but it shows how
> pigments can be used to create the shape of distant thunderclouds. It's all one
> pigment, generated with the 'pigment control trick' and MegaPov's spline.
> 

That already looks quite promising for some kind of near horizon cumulus
clouds.  It only will be quite difficult to achieve realism on both near the sun
and on the opposite side.  

Christoph

--
Christoph Hormann <chr### [at] gmxde>
Homepage: http://www.schunter.etc.tu-bs.de/~chris/

Post a reply to this message

Simon de Vet wrote:
--snip--
> 
> I've looked at other people's cloud techniques, and they all use 3-d to generate
> 3-d looking clouds. Some use media, or isosurfaces, or stacked planes, but all use
> actuall objects.
>
--snip--

If I may beat my own drum for a sec, I posted a non-media 2D cloud
example not too far back. The method provided for some 3D lighting
effect. For the results, the rendering times were pretty fast. I'll take
the liberty of sticking in another example my pigment clouds below.
(This was from an effort to copy the technique from the TGLTLSBFSSP page
which you mentioned).

I have always thought that simple bump mapping should do the trick of
providing the correct shadows to pigment. So far though, I have had very
little success with it in POV. Not sure why ...

Abe

Post a reply to this message

Attachments:
Download 'clouds_simple.jpg' (33 KB)

Preview of image 'clouds_simple.jpg'

Regarding the clouds at Hugo Elias' page, he _does_ apply cloud shading 
via bump mapping in the snippet of pseudo-code at the end. You can also
see it in the example image.

Abe

Post a reply to this message

"Paul Vanukoff" <van### [at] primenetcom> wrote in message
news:39a71fe1@news.povray.org...
>
> AWSOME ROLEX! even.

I guess I should post another of my stern warnings against posting photos
and trying to pass them off as POV.

[ ;-) Good work]

--
Eric
--------------------
http://www.datasync.com/~ericfree
--------------------
"I don't like it, and I'm sorry I ever had anything to do with it."
- Erwin Schrodinger talking about Quantum Mechanics.

Post a reply to this message

Abe wrote:
> 
> Regarding the clouds at Hugo Elias' page, he _does_ apply cloud shading
> via bump mapping in the snippet of pseudo-code at the end. You can also
> see it in the example image.
> 
> Abe

Seems you are right, but the effect is not very strong.  I think simple bump
maps probably are not sufficient in many cases like this:

http://www.wolkenatlas.de/wolken/wo12625.htm

Of course these are somewhat extreme, but the general type of those clouds is
very common.  

Christoph

--
Christoph Hormann <chr### [at] gmxde>
Homepage: http://www.schunter.etc.tu-bs.de/~chris/

Post a reply to this message

Christoph Hormann wrote:
> 
> Seems you are right, but the effect is not very strong.  I think simple bump
> maps probably are not sufficient in many cases like this:
> 
> http://www.wolkenatlas.de/wolken/wo12625.htm

An understatement.

> 
> Of course these are somewhat extreme, but the general type of those clouds is
> very common.

...and very desirable ;)

Abe

Post a reply to this message

Why are image maps so frowned upon in the POV community?  Almost all of
the big wig renderers use them solely.

GrimDude wrote:
> 
> Image_map and IRTC? Bite your tongue!
> 
> Grim

Post a reply to this message

Simon de Vet <sde### [at] istarca> wrote:
> I've looked at other people's cloud techniques, and they all use 3-d to
generate
> 3-d looking clouds. Some use media, or isosurfaces, or stacked planes, but
all use
> actuall objects.
[snip]
> I'm attaching a test image I made before I gave up on this technique, due
to time
> limitations. It's a very early image, with no pseudo-3-d effects, but it
shows how
> pigments can be used to create the shape of distant thunderclouds. It's
all one
> pigment, generated with the 'pigment control trick' and MegaPov's spline.

Yes, I think this method does hold a lot of promise.  A while ago I put
together a few animations of various types of clouds, which do include
sun-shading (see the Current Projects page at
http://www.geocities.com/ccolefax/projects.html).  The code is all quite
basic POV 3.1, and I've parametised most aspects of the clouds.  I think the
trick is to find pigments that look effective at high resolutions...

Post a reply to this message