On Saturday 26 March 2011 10:16:47 thorsten.i.r...@jyu.fi wrote:
> > How about using plain "old" textured plane crosses for clouds at bigger
> > distances? (the way trees are done, only with a (or several) horizontal
> > plane(s) added)
> > These might work better for bigger structures like cumulonimbus, and
> > such,
> > too, as you can "control" the shape a bit. (no fancy shader needed ;) )
>
> I don't think that works.
>
> Issue 1: I have never been able to work out a geometry for a static,
> believable cloud. I have tried plane crosses, intersecting spheres and
> ellipsoides, stacks of curved sheets, bubble foam geometries,... no joy,
> all looked quite bad even from some distance. I have tried for a month, no
> success. However, you are one of the best 3d modellers around here - so
> maybe you can pull it off.
>
> Issue 2: Shading - if faraway clouds are not shaded, they look unrealistic
> if nearby clouds are shaded away from the sun. If they are shaded, I don't
> see how the shading of any static geometry will be able to mimick the
> shading of the rotated sheets - especially plane crosses look very
> different when shaded.
>
> Issue 3: Transitions approaching the cloud: Nearby clouds are created as
> random stacks of different cloudlets - so there's no way one could know in
> advance how the detailed version of the cloud will look. Therefore, the
> faraway proxy and the nearby realistic cloud will almost certainly look
> differently, and this will give you a rather ugly transition.
>
> Issue 4: Transitions leaving the cloud: Nearly as bad - once in the
> scenery, I make no distinction of what a cloud is, individual textures are
> on their own - in principle a cloud can partially decay, merge with a
> different one, merge with a layer, separate itself from a layer,... That
> goes along with nature where 'cloud' is also not a well-defined object. So
> in going from a multi-texture stack to a single static model, you'd
> somehow group clouds again and make a decision which of your textured
> surfaces are supposed to represented by the placeholder. Consider an
> undulatus pattern - what is it you'd replace - the whole layer, one strand
> of clouds, parts of a strand where it is disconnected? How do you group
> technically - you'd somehow probe the whole geometry and match it to some
> predefined pattern. You also can't store the info what a cloud is supposed
> to be at creation time, because clouds are allowed to evolve and to
> change.
>
> So, looking into the details, it's really far from trivial how to use
> placeholders, and that is why I haven't been able to work out a viable
> scheme.
I'll think about these issues, and see if and what I can come up with.
> > I'll have a go at retexturing, redimensioning and choping this weekend if
> > that's ok with you.
>
> Yes, certainly. I suggest that we pack your /Models/Weather/ then into a
> tarball and I host that for the time being as a dds texture patch to Local
> Weather so that people can test what works better for them. Once we know
> how this affects different systems, we can decide what to do and which
> version should go into GIT.
>
> Based on the Electra responses in the forum, dds doesn't seem to run for
> everyone without problems...
>
> * Thorsten
Theoretically, clouds wouldn't be affected by those issues, as those are
related to the normalmap format (it's a newer compression method, and it's
task specific).
------------------------------------------------------------------------------------------------------
That could come up in the future, if things improve, maybe we'll use
normalmaps on the clouds too, to enhance them, but that's another story.
------------------------------------------------------------------------------------------------------
I've started in advance with the chopping (to give me a head-start ;) ), and
one conclusion I can draw, from what I've seen so far, is that texture usage
is very inefficient in some cases. So i think we'll at least free up some of
the
memory used regardless of the final format of the textures will be in. (There
are textures with >60% empty space in them, that's wasting video memory, as
these get decompressed into videoram; I'm not criticizing your work, which i
can only appreciate even more now seeing what a huge task it was if only to
model and texture all the clouds, which I imagine was a minor part in the
development).
On that note, I take this opportunity to thank you again for all the
development done on this subject, and I'm glad I can contribute in some way to
it.
Have a nice springtime week-end (at least here it's shaping up that way
weather wise :) ),
Emilian
------------------------------------------------------------------------------
Enable your software for Intel(R) Active Management Technology to meet the
growing manageability and security demands of your customers. Businesses
are taking advantage of Intel(R) vPro (TM) technology - will your software
be a part of the solution? Download the Intel(R) Manageability Checker
today! http://p.sf.net/sfu/intel-dev2devmar
_______________________________________________
Flightgear-devel mailing list
Flightgear-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/flightgear-devel
