Hi

Great work Jaime.

Only one comment, the trees seem too similar both in shape, variety of
species and distribution. But hell that's just nit picking, it's superb :-)

Mick

Post a reply to this message

de news: 40e7d173@news.povray.org...

Wow
now I know where I'll spend my trekking holydays.
I love the stone structure showing tilted sedimentary rocks layers and
vertically bound erosion.
Just my piece of nitpick: you have (at least) two cloud types: cumulus and
altostratus.
Altostrati should be way higher than cumuli and faded by atmospheric haze
down to horizon.
But it is just in a realistic aiming view, your clouds look as on a
classical painting :-)

Marc

Post a reply to this message

Christoph Hormann wrote:
> Nice, how does the render speed of the sky part compare to the terrain? 
>  I assume the sky is quite a bit slower?

   Yes... although not much more. A solution I figured out for quick 
skies is to choose a fixed sun placement and render a spherical map of 
the media clouds to use it later on the scene with the same sun 
position, but it needs very high resolution images. Rendering a decent 
skymap can take several weeks... and it only serves for a concrete sun 
placement.

   BTW, I used 3 intervals, not 2 as I said before.

> Is this randomly set up or did you select the composition manually?

   Almost totally random, as I added an optional main seed to generate 
individual seeds for each feature. In this one I've manually selected 
the terrain and fog colors, the sun placement, and the diameter and 
center of the vegetation area (based on the camera placement, which is 
random, as it should be!;)

   There is something very addictive in trying seed after seed until you 
find a nice setup... I'm using the animation features to get a different 
landscape on each frame, so I can let it automatically render many low 
res tests, to latter chose the best ones and tweak them manually.

--
Jaime

Post a reply to this message

Jaime Vives Piqueres wrote:
>   Yes... although not much more. A solution I figured out for quick 
> skies is to choose a fixed sun placement and render a spherical map of 
> the media clouds to use it later on the scene with the same sun 
> position, but it needs very high resolution images. Rendering a decent 
> skymap can take several weeks... and it only serves for a concrete sun 
> placement.

But when you have one sky map you can rotate it arbitrarily around the 
vertical axis to get different sun positions in azimuth.  The height of 
the sun is of course fixed, as well as the position of the clouds 
relative to the sun.  And for good results you would of course need a 
high dynamic range map.

> 
>> Is this randomly set up or did you select the composition manually?
> 
> 
>   Almost totally random, as I added an optional main seed to generate 
> individual seeds for each feature. In this one I've manually selected 
> the terrain and fog colors, the sun placement, and the diameter and 
> center of the vegetation area (based on the camera placement, which is 
> random, as it should be!;)
> 
>   There is something very addictive in trying seed after seed until you 
> find a nice setup... I'm using the animation features to get a different 
> landscape on each frame, so I can let it automatically render many low 
> res tests, to latter chose the best ones and tweak them manually.

This is exactly what i meant.  In fact it does not matter much if you 
manually tweak the settings or if you try random scenes until you find 
one that looks good.  The thing i am having trouble with is to get the 
random parameter selection to *always* produce a scene that looks 
reasonable.

Christoph

-- 
POV-Ray tutorials, include files, Sim-POV,
HCR-Edit and more: http://www.tu-bs.de/~y0013390/
Last updated 01 May. 2004 _____./\/^>_*_<^\/\.______

Post a reply to this message

Jim Charter wrote:
> Words will surely fail...

   Specially due to my poor english skills...

> But among other things, I love the complexity, but with variation, that 
>  you get in the vegetated, eroded slope descending on the left,...I can 
> see getting one or the other, but both? And then you add the further 
> element of exposed rock structure on the vertical face!

   Multiplying two pigment functions gives fantastic complexity with 
great variation, specially if the two pigments are at different scales.

> ...ability with complexity to produce the look of landscape.  And 
> contemplating the look of a landscape is exactly what we do when we 
> climb to a vantage and gaze out.

   Yes, I was going for a sort of impressionism, giving just enough 
detail to let the brain do the rest. It requires a proper resolution, 
not too low to avoid losing details, but also not high enough to show 
the real CG nature.

> ...  But the variation you produce looks truly various not 
> simulated, and just in the way a landscape does. How do you achieve 
> that, one wonders. To automate you must establish parameters, and 
> general structures, to direct random events. The additions of scene 
> elements can never rise above a manual checklist. So landscape is not 
> infinitely various? Yet it is this sense of unlimited variation that 
> gives so much pleasure.  

   It is infinitely various in the sense that you must provide the 
isosurface function. This function is later used with a random color map 
and translated and turbulated randomly, so the variation you can get is 
really high.

 > Is your guide post creation itself, or the look of creation?

   I'm not sure to understand correctly the question... if you ask for 
my motivation, it is always to be surprised by the results, and this is 
greatly accomplished with highly random scenes like this one. I even use 
randomly guided cameras and lights placement on my "usual" scenes, as my 
composition abilities are not well developed (but I can recognize a good 
composition when I see it, which is enough with trial&error techniques).

--
Jaime

Post a reply to this message

Mick Hazelgrove wrote:
> Great work Jaime.

   Thanks!

> Only one comment, the trees seem too similar both in shape, variety of
> species and distribution. But hell that's just nit picking, it's superb :-)

   Yes, there is only one object, scaled and rotated randomly. Next step 
with vegetation will be to add support for user-supplied objects, as 
well as giving some predefined ones. Hmmm... I'm thinking now I can use 
eval_pigment for the distribution/mix of the species too. Thanks!

--
Jaime

Post a reply to this message

Marc Jacquier wrote:
> Just my piece of nitpick: you have (at least) two cloud types: cumulus and
> altostratus.

   Yes, I basically tried two types of clouds: cumulus and high altitude 
clouds.

> Altostrati should be way higher than cumuli and faded by atmospheric haze
> down to horizon.

  In fact the high layer is really much higher than the cumulus, but it 
seems not very noticeable. Perhaps the scale of the pattern has 
something to do with it...

> But it is just in a realistic aiming view, your clouds look as on a
> classical painting :-)

   Thanks!

--
Jaime

Post a reply to this message

Christoph Hormann wrote:
> But when you have one sky map you can rotate it arbitrarily around the 
> vertical axis to get different sun positions in azimuth. 

   Yes, I did not think on it, but you are as usual right.

> And for good results you would of course need a 
> high dynamic range map.

   Not necessarily, I think, only if you want very accurate radiosity. 
For me the biggest problem is the rendering time required to create an 
extensive collection of cloudscapes.

> The thing i am having trouble with is to get the 
> random parameter selection to *always* produce a scene that looks 
> reasonable.

   I think you are way too exigent... but I do not doubt you will figure 
out a solution. :)

--
Jaime

Post a reply to this message

Jaime Vives Piqueres wrote:
> 
>> And for good results you would of course need a high dynamic range map.
> 
> 
>   Not necessarily, I think, only if you want very accurate radiosity. 

For reflections (water) this can be important as well although you can 
trick quite a lot with reflection exponent.

> For me the biggest problem is the rendering time required to create an 
> extensive collection of cloudscapes.

I think for high resolution renders (which your recent scenes seem very 
suited for) this is probably not a solution anyway.

>> The thing i am having trouble with is to get the random parameter 
>> selection to *always* produce a scene that looks reasonable.
> 
> 
>   I think you are way too exigent... but I do not doubt you will figure 
> out a solution. :)

I am quite near to start my new LOTW-2 but i simply still get quite a 
lot of 'boring' scenes.  Since i plan to do 20 tests for every 
generation of which only 4 get selected this is not such a big problem 
though.

Christoph

-- 
POV-Ray tutorials, include files, Sim-POV,
HCR-Edit and more: http://www.tu-bs.de/~y0013390/
Last updated 01 May. 2004 _____./\/^>_*_<^\/\.______

Post a reply to this message

Jaime Vives Piqueres wrote:

> 
>  > Is your guide post creation itself, or the look of creation?
> 
>   I'm not sure to understand correctly the question... if you ask for my 
> motivation, it is always to be surprised by the results, and this is 
> greatly accomplished with highly random scenes like this one. I even use 
> randomly guided cameras and lights placement on my "usual" scenes, as my 
> composition abilities are not well developed (but I can recognize a good 
> composition when I see it, which is enough with trial&error techniques).
> 
I'm not sure I understand the question either.  But you turned it into 
an understandable and interesting answer.  I think that what I was 
reaching for was somewhat rhetorical.  To wonder about what it really is 
that we find so pleasurable about landscape, and how that relates to 
your process.

Post a reply to this message