PDA

View Full Version : Hypervoxels moving clouds with Dpont sunsky-Revisit part2.



prometheus
04-21-2014, 02:39 PM
Well...since I got my desktop up an running now, and springtime is here now in sweden..which means not only constant grey sky coverage, but in fact some of the more interesting cloud formations when the air starts to warm up and clouds forming..I felt inspired to get back to testing hypervoxels again.

Itīs not the best or perfect solution, but I sort of like to fiddle with it..rather than skytracer..the other plugin stuff is additional cost for ozone etc...and hdr animated photos isnīt what I want here, I want to control Everything from shadows casts on the ground, to the speed of clouds, fluctuation, cloud shadow density and forming,light interaction and angles etc...so it is the painful long rendertime approach:)
Skytracer is out of the question, general sky looks worse than dpons sky, the clouds are not fully volumetric and canīt cast shadows etc, so please drop that and invest in a fog environment compatible with
sunsky and a true volumetric sun, the sunsky sunlight and native lightwave domelights canīt produce that, and the only other option is distant or point lights ..but it isnīt good enough.

the video sample was completly VPR draft mode rendered, thus this clip shows some stuttering, which I canīt tell the cause of, either the vpr fps setting or some other refinement settings? this is of course not how you should render out in the end, but render to tga or png for example.

VPR provided the rendered sequence at aprox 46 seconds or something, while a final render reached 7 minutes with radiosity too, so I would have needed to render over the whole day or more for the whole sequence, but I wonīt be doing that until I am satisfied ..still have a lot to work on if I am up too it that is.

You can check either on vimeo or youtube, I think the quality is better on vimeo though, due to the immense time needed for rendering animated clips with volumetrics, the clip is as usal ..too short:) if you ask for more, please sponsor me with a renderfarm or something:)


https://vimeo.com/92539727

https://www.youtube.com/watch?v=N3i7v1b7kSQ

- - - Updated - - -

Description...
well...here we go again, more cloud testing with particles and hypervoxels, I have to mention that this was only a
VPR draft renderer ..otherwise it would have taking to long time for initial testing, the VPR makes it at aprox 46 seconds per frame, 5000 particles, hypervoxels,fog.
a final render would probably take around 7 minutes, but that also with radiosity.

This draft vpr render has some stuttering due to vpr inefficiency when refining or the vpr fps settings, and it also has some flickerings due to local density I think, but I wanted that to smooth out hypervoxels a little.
The quality of the hypervoxels is medium, good and even very good would make it better of course, but also way longer to render.

particles ar moving at 35 meters per second (a little to fast here for my taste) in the particles velocity motion... the hypertexture is turbulence, and the hypertexture effect is velocity translate, that gives a fluctuation and a transform of the texture/cloud, but I do recognize that it isnīt good enough, since that is based on each particle.
Instead I would also need to complement that with some changing of the birth rate texture I have in the particles birth rate tab, thatīs where the overall particle density is set.
No post processing in the clip was done.
the clip and the images differs, due to no radiosity in the clip (darker ground shadows) I could also have used a domelight/skylight instead.
the noise in the clip is also evident on the ground due to vpr not refining good enough in draft mode.

prometheus
04-21-2014, 02:47 PM
and hereīs some settings..
No birth rate textures gives no holes in the clouds, but you would rather have it the other way around...mostly. second image is without the birth rate texture that controls the overal particle/cloud density.

http://forums.newtek.com/attachment.php?attachmentid=121559&d=1398113061


http://forums.newtek.com/attachment.php?attachmentid=121560&d=1398113093


http://forums.newtek.com/attachment.php?attachmentid=121561&d=1398113117


121559 121560 121561

prometheus
04-21-2014, 02:55 PM
Ohh..I actually posted this in the wrong section, if chuck sees this, please feel free to move it to the lightwave general tips and tricks section.

Ztreem
04-21-2014, 04:03 PM
I've not read through the entire post but I looked at the youtube link and on my iphone it looks really good, great work.

prometheus
04-21-2014, 04:12 PM
I've not read through the entire post but I looked at the youtube link and on my iphone it looks really good, great work.

wow..and thatīs on a small iphone and with stuttering frames from vpr and youtube poor quality,and me not optimizing the clip either for youtube, thatīs good to know, I might actually do a proper render then later on.
Thanks for the feedback.

Ztreem
04-21-2014, 04:37 PM
I'll check it out at work tomorrow with my regular computer and maybe I'll take everything back then. :D

Hail
04-21-2014, 04:58 PM
Great tips!
Subscribed!:)

spherical
04-21-2014, 10:45 PM
Really nice feeling in those. Well done. I'll have to try to figure out what you're doing through the settings images, 'cuz I just gotta know!

CaptainMarlowe
04-22-2014, 02:19 AM
Nice work. Many thanks for sharing the settings !

Danner
04-22-2014, 06:00 AM
I didn't think it was possible to get clouds to look this good with LW alone. Already looking awesome, keep us posted.

Ztreem
04-22-2014, 07:20 AM
Checked the Vimeo clip on a real screen today and it still looks really good.

gerry_g
04-22-2014, 07:43 AM
the way the procedural moves through the volume is a dead give-away and an indication of the limitations of the process but for all that it looks pretty amazing and way better than I can manage

Svenart
04-22-2014, 09:58 AM
Iīm a little bit surprised that it is possible to do this in LW. I Think it could be a nice alternative to other solutions. But Iīm not really sure if it looks that nice in hd. Also the rendertime would be probably a bit high in 1920x1080. But still very good!

prometheus
04-22-2014, 10:34 AM
the way the procedural moves through the volume is a dead give-away and an indication of the limitations of the process but for all that it looks pretty amazing and way better than I can manage

well youré right and wrong, the procedural donīt move through the volume properly:) I think that is actually what you ment, itīs the hypertexture effect only on a per point voxel..but you are also right:)
so yes that is what I mentioned in the description of how the velocity hypertexture effect affects it on a per point voxel, and not through the whole volume, I could test with world coordinates...but I donīt think that looks good, the fluctation/evaporation effect needs to be done more in the overall cloud density, and for such I would have to animate the particle birth rate density by moving that texture a little, but havenīt come that far yet.

If hypervoxels in the future could use very large planes and implemented on a per object basis and still having the option to use a velocity translate effect when setting cloud movement speed, that would be the way to goo, sort of similar to how ozone does it, but with infinite procedural volumetric cloud planes, equally with ogo taiki you have infinite procedural volumetric cloud planes, but with ogo taiki you donīt have a natural control so to speak, you have to animate the cloud density texture for it or scale it, unfortunatly ogo taiki wonīt work with 64 as I know of, and is so messy to set up from scratch and the quality is hard to get right too, and if you do...long long rendertimes.

I could tune down the hypertexture effect speed, if that might look better, initial speed testing was to funny..had it on too much and it just flickered all over the place, the effect here should only be a small amount of speed I think, and
as I mentioned before I would need to move the actual birth rate texture in y axis or maybe scaling it, or using envelopes or other animatable procedurals, but this must be done delicatly and not too much.
Another way I would like to try out some time, a simple fluid simulation rendered in top view or front view ortho, then use that as density distributor and to drive the motion for the hypervoxels particles later on, so I get the movement simulated from fluids, but I can get all the shadow/lighting and hypertexture noise from hypervoxels.



Michael

prometheus
04-22-2014, 10:37 AM
Iīm a little bit surprised that it is possible to do this in LW. I Think it could be a nice alternative to other solutions. But Iīm not really sure if it looks that nice in hd. Also the rendertime would be probably a bit high in 1920x1080. But still very good!

If the Lightwave group checkīs all my threads about how I think hypervoxels need to be improved, and preferably actually can implement it, we can expect to see much more realism, even though I think I can push this some steps more to realism.

Rendertimes will probably be very long in any application doing volumetrics, that resolution of 856x492 was around a little over 7 minutes..With radiosity..which you could exclude and use some occlusion instead, cam AA at
2x2 a little to low for use with objects, instanced grass etc.

If I were to throw in a dense subdiv mesh for terrain, and instances for trees..all that needs freezing and AA so ..long rendering times, and you just canīt expect to pull any longer animation frames out of it with one single home computer as I have.

some small stretching amount, using the right particle size, and in this case very large value of random scale, you donīt see uniform scaled puffs in real clouds that often..so that is one of the tricks, thou if we had a procedural cloud plane implemented, you can skip such tricks and only rely on the texture itself.

Michael

prometheus
04-22-2014, 10:53 AM
I will show how the fog affects the look of the clouds too...later on.

zapper1998
04-22-2014, 11:01 AM
kewlness

kadri
04-22-2014, 09:33 PM
Looks nice.

prometheus
04-23-2014, 01:25 AM
Hereīs some post process fiddling, not exactly from the same scene, I made some other changes in the clouds with some other voxel settings etc...here I just want to show some postprocessing in photoshop.

so the first image is a straight render, the second image Is blurred a little on the clouds, maybe to much by the way..and I did some color correction... I also made a faked godray layer with the radial zoom trick, and the second image is a blend of 80 percent opacity over the first straight rendered image.
The third image is the same as the second but no blending with the first layer, thus it has 100% of the effects.

Godrays with volumetric lights only in lightwave isnīt a joy exactly in terms of setting up ..and rendering, I have done that before with some particle fields and volumetric lights ..It looked decent, but this scene I just couldnīt get any decent out of that...it simply halted everything to sleep, so ozone still shine with that..and maybe ogo taiki.

http://forums.newtek.com/attachment.php?attachmentid=121586&d=1398237571

http://forums.newtek.com/attachment.php?attachmentid=121587&d=1398237587

http://forums.newtek.com/attachment.php?attachmentid=121588&d=1398237611



121586 121587 121588
Michael

Waves of light
04-23-2014, 09:34 AM
This is excellent Michael, thanks for sharing.

04-23-2014, 12:37 PM
Oh boy, this is some nice stuff you are sharing, Michael. Glad you got your computer fixed.

Looking forward to all else you might share. Ha, I'll even ask for one: got a scene to share?

prometheus
04-23-2014, 01:23 PM
Oh boy, this is some nice stuff you are sharing, Michael. Glad you got your computer fixed.

Looking forward to all else you might share. Ha, I'll even ask for one: got a scene to share?

Hi Robert!
Im just sharing some screencaps and descriptions for now, and if you create the scenefiles later for yourself, it will be your work and your pride:)
Feel free to ask anything you like about it, if you stumble on obstacles in the process, keep in mind that it isnīt all about understand the particles and hypervoxels solely, you need to
check the real world and study clouds etc, and it helps having an artistic eye too, if not..it will be much harder, hypervoxels and particles is only your tool brushes.

I have some more descriptions to explain and share later on, these things takes time to tweak though, and my awake hours went long over 16 hours causing me to postpone the natural day rythm unfortunatly, so I might need to give it a rest for a while and go out for a walk and study the real clouds a little more.

Michael

H_Molla
04-23-2014, 02:30 PM
very nice..

gerardstrada
04-25-2014, 07:41 PM
Indeed. Interesting tests. Your results look more like Helios' clouds (http://www.quadspinner.com/Helios/Features) than to Vue/Ozone's clouds.

http://www.quadspinner.com/img/Helios/2_conversion.jpg

mainly with the blur trick...



Gerardo

prometheus
04-25-2014, 07:58 PM
Indeed. Interesting tests. Your results look more like Helios' clouds (http://www.quadspinner.com/Helios/Features) than to Vue/Ozone's clouds.

http://www.quadspinner.com/img/Helios/2_conversion.jpg

mainly with the blur trick...



Gerardo

yeah you know, the shadow and lighting of voxels arent that bad, if you know how to delicatly tweak and and adjust shadow strength, I usually go for a very low shadow setting, sometimes as little as 0,5-2 depending on thickness,opacity and density, if you raise opacity you block out light, lower means more penetration, same with thickness etc, if one learns how to combine all those settings and balance it, you can get very far with it, also a matter of slight ambient light sometimes.

the helios cloud on the right you refer to has the new subvapor scattering algoritms to make it softer and more full in volume as well.
I hope there will be a demo available that can work with a ple version..I just got to try it out.
I do hope though that that feature is controllable, I would like to tweak the scattering to be not as full but not as flat as the first image.
I think modo also has some sort of rayleigh forward scattering control with more light penetration control..
Houdini has itīs light propagation control.

textured shadows can also help with that in voxels, but I rarely use them, they can give fuller volume textures but it can also give to much texture that means the overall volume look getīs destroyed...that is because the texture itself is based on a per point/particle basism, so sometimes you can sort of fake it out by not having to much shadow and also raise ambience lighting in the voxels so the texture shadow donīt get too defined.

So the look of the image to the right is doable in terms of getting more fuller volume and softer, but it would mean it also adapts the texture per point basis as mentioned
and it might not always look good, in helios the cloud density is purely procedural on a cloud volume..not based on a particle spherical size..thus it might look better.

we have a couple of major issues with hypervoxels...
the falloff isnīt working, donīt think it ever has? strange though...it works if you use the procedural editor hack and apply hypertextures there, from there the falloff
works..but then again itīs a pain to tweak since that node hack do not update in viper or VPR.

the thickness texture channel do not respond to any local density gradients, so softening has to be done with dissolve..but neither gives a really good smooth edge..
for that we need a density curve in either density channel or maybe a curve controller in smooth and thickness channels.

Then itīs a matter of the actual hypertexture itself....it goes out in a spherical pattern build up for the procedural texture... and sometimes that gives very hard cutoff edges...this is known also in other software and I think Helios has a new solution to that..even though the metaclouds worked decently.


Michael

chikega
04-25-2014, 08:10 PM
Some very nice results! In addition to updating Hypervoxels, a true physical sky with settings for turbidity, ground albedo, etc... would be most welcome.

http://www.amaanakram.com/?page_id=290

prometheus
04-25-2014, 08:20 PM
Some very nice results! In addition to updating Hypervoxels, a true physical sky with settings for turbidity, ground albedo, etc... would be most welcome.

http://www.amaanakram.com/?page_id=290

ahh..you mean a full volumetric atmosphere? that I want to...dpont sunsky isnīt there...at least not to match atmospheric phenomena in the same way, and as I understand it, denis donīt have the resources to work on such things, so not even a new suntype that is volumetric..then again I would think it must have a volumetric fog layer, not adapting a volumetric handler to the light itself, it would be wonky to get true sun scale settings otherwise...so something like what ogo taiki has when setting fog-layer, but easier to use...faster and with a spectral sky atmosphere..not that mess of
filters and settings for the sky as it is in ogo taiki.

I have some fears though that we will not see anything of some improvements on this at all for lightwave 12...just a hunch...I suspect they will focus on other things.

Michael

ary3d
04-25-2014, 09:06 PM
Impressive test Michael!

I would like to see a check option to make the 3d texture as one entity, and not per particle sphere boundary, I mean you have a cloud made of particles, an each one has his size and fall off, but it repeat the 3d texture in each particle and make a pattern, so, and option like "blend volume particles" will be cool.

I don't know if I made my self clear! ;D

greetings, David

gerardstrada
04-25-2014, 09:35 PM
Some very nice results! In addition to updating Hypervoxels, a true physical sky with settings for turbidity, ground albedo, etc... would be most welcome.

http://www.amaanakram.com/?page_id=290
DP Sunsky is a true physical sky with settings for turbidity, backscattered light, aerosol, ozone, water vapor, etc. Last sunsky model have also high atmospheric turbidity for much more realistic sunsets and an "Albedo" ground influence. We have also Sunsky Atmosphere, to add there sun flare in the image (varying with angle of view, sun position, Mie scattering, Rayleigh scattering, Light Inscatter, Light Attenuation...). Both models in DP Sunsky are based in the same models for Arnold (the native) Hosek&Wilkie, which has more artist-friendly (read simple) parameters and more realistic solar disc with the limb darkening effect. And the Preetham model (same model the plugin you have linked is based on). Also, wavelengths in DP Sunsky (Preetham sky) are integrated from 380nm to 780nm at 5nm intervals, using CIE 2-degree observer luminance functions, which is a more accurate color reproduction than the plugin you have linked for Arnold. And similar thing for Hosek&Wilkie model in DP Sunsky. Moreover, DP Sunsky is the only physical sky solution that provides adapted colors to display (none of those models for Arnold, Vray, MentalRay or even Vue have this proper correction). And even more, we have there, calculations in scene-referred color spaces like ACE (from AMPAS) and RIMM (from ICC). If you ask me, DP Sunsky is much much better.



Gerardo

shrox
04-25-2014, 10:14 PM
I have been getting some good clouds and some rolling smoke out of HVs.

prometheus
04-26-2014, 09:19 AM
Gerardo[/QUOTE]




DP Sunsky is a true physical sky with settings for turbidity, backscattered light, aerosol, ozone, water vapor, etc. Last sunsky model have also high atmospheric turbidity for much more realistic sunsets and an "Albedo" ground influence. We have also Sunsky Atmosphere, to add there sun flare in the image (varying with angle of view, sun position, Mie scattering, Rayleigh scattering, Light Inscatter, Light Attenuation...). Both models in DP Sunsky are based in the same models for Arnold (the native) Hosek&Wilkie, which has more artist-friendly (read simple) parameters and more realistic solar disc with the limb darkening effect. And the Preetham model (same model the plugin you have linked is based on). Also, wavelengths in DP Sunsky (Preetham sky) are integrated from 380nm to 780nm at 5nm intervals, using CIE 2-degree observer luminance functions, which is a more accurate color reproduction than the plugin you have linked for Arnold. And similar thing for Hosek&Wilkie model in DP Sunsky. Moreover, DP Sunsky is the only physical sky solution that provides adapted colors to display (none of those models for Arnold, Vray, MentalRay or even Vue have this proper correction). And even more, we have there, calculations in scene-referred color spaces like ACE (from AMPAS) and RIMM (from ICC). If you ask me, DP Sunsky is much much better.



Gerardo
think we have discussed that earlier..yes it is a physical sky, but I think he ment true physical sky with the inclusion of fog/haze effect as well as air scattering in full volumetrics, I do not see any signs of such scattering/spectral model that goes in depth or volumetric..such as in vue for instance.

Itīs a great tool to have though and gives nice cloudless skies and I am thankful towards Denis for providing it, and he also helped and actually fixed shawow casting from hypervoxels upon my request very quickly.
Even though we have turbidy and albedo control, it donīt affect the skies in volumetric sense it seems, itīs not a fog model that goes volumetric and you can place any object at any distance..the turbidity will not be seen to affect that object..which in a physical sky I think it would to some degree.

I use the native fog to both color and distance fade the clouds in to the sky to get them to look more as they do in the real world, and I check the use background color for it..and as such it picks up the settings from sunsky and impose it over the hv clouds.

If you have some object in some distances ....this will cause some issues where the horizont will impose over those object as the sky backdrop is selected as fog color, this will render a very sharp edge and a color change over that object and if you have the sun facing you, the sun disc will appear in front of those objects...and thatīs not how it should be.

A physical sky has fog and haze and it scatters light anisotrophicly in the air depending on that fog and turbidity, so we donīt have a full physical sky, I do not seperate the sky from atmosphere here when naming it in terms.
I also think we lack a sun decay in this model..but that is also a volumetric atmospheric effect I suppose.
The atmospheric script he has ..just gives some glares around objects on the ground etc and some additional glare around the sun, but it isnīt near the type
of atmospheric look you get from vue.

Sunsky gives nice clean air sky so to speak, and even some foggier turbidity looks that is good looking, that is why I would want a mix of sunsky with the atmospherics of ogo taiki and also have ogos textured based procedural cloud layers and fog/volumetric layers for godrays and such.
the color attenuation affects hypervoxels pretty nicely and hypervoxels picks up colors and lights nicely..so it looks realistic, so in that sense it donīt seem to need much work..

however ..when the sunligh is in itīs highest peak and scatters light through the atmosphere, the global illumination also affects clouds much more, for this I fake it with hypervoxels global light illumination..but that means that light will still be as strong when you lower your sun, so in such case you need to reduce the ambient lighting because the clouds should appear darker in how much light they pick up in the inner volume in such cases, you could of course map that color to the sunsky I think in the envelopes.

A good thing about sunsky, you just apply it and change the sun or turbidity and it looks quite naturalistic, as opposed to ogo taiki where you have to mix and fiddle with to many settings to get it look right, when you do get that right it ogo taiki..it looks awesome with all the volumetrics...but slow...we just canīt have the cake and eat it at the same time it seems.

prometheus
04-26-2014, 09:51 AM
Impressive test Michael!

I would like to see a check option to make the 3d texture as one entity, and not per particle sphere boundary, I mean you have a cloud made of particles, an each one has his size and fall off, but it repeat the 3d texture in each particle and make a pattern, so, and option like "blend volume particles" will be cool.

I don't know if I made my self clear! ;D

greetings, David

Yeah..I think I know what you mean, using world coordinates sort of seem to do that a little, but I havenīt been able to do any good motion with that option, you do not get the texture to follow the particle movement properly, which you would have to assign to some other reference object I think...anyway, I havenīt managed to solve that.

Hypervoxels now do blend "volume particles" introduced from 11.5 or 11.6 I think, but that blending option doesnīt do any good for this, I think they failed on implementing
a good blend here, the old dynamite voxel engine had this smooth blending tension ..just like they are in the surface mode, then again it was slow.

modo also has a much better blending between the voxels as to what I have seen, recent showcases of clouds from modo is based on a volumetric container though so you just pull the texture to move, modo also seem to have smoother edges in the softness of the voxels, and it has some more options to control forward scattering.
strangely though..I havenīt seen so much in the gallieries from modo users..and not many of them good either...the tech is maybe just half the equation and they
might need someone good at it.
So you might wanīt to try it david:) Im to obnoxius diving in to some policies around their demo trial of 15 days and paying for another version for 15 days more etc.

a blend function or volume calculation based on particle amount or particle density to make those the bounding container where the whole volume is applied on, that would be nice .. I think that is maybe what you mean? and the following hypertexture is shaping around that and not around each individual particle.

gerardstrada
04-26-2014, 06:54 PM
think we have discussed that earlier..yes it is a physical sky, but I think he ment true physical sky with the inclusion of fog/haze effect as well as air scattering in full volumetrics, I do not see any signs of such scattering/spectral model that goes in depth or volumetric..such as in vue for instance.

Itīs a great tool to have though and gives nice cloudless skies and I am thankful towards Denis for providing it, and he also helped and actually fixed shawow casting from hypervoxels upon my request very quickly.
Even though we have turbidy and albedo control, it donīt affect the skies in volumetric sense it seems, itīs not a fog model that goes volumetric and you can place any object at any distance..the turbidity will not be seen to affect that object..which in a physical sky I think it would to some degree.

I use the native fog to both color and distance fade the clouds in to the sky to get them to look more as they do in the real world, and I check the use background color for it..and as such it picks up the settings from sunsky and impose it over the hv clouds.

If you have some object in some distances ....this will cause some issues where the horizont will impose over those object as the sky backdrop is selected as fog color, this will render a very sharp edge and a color change over that object and if you have the sun facing you, the sun disc will appear in front of those objects...and thatīs not how it should be.

A physical sky has fog and haze and it scatters light anisotrophicly in the air depending on that fog and turbidity, so we donīt have a full physical sky, I do not seperate the sky from atmosphere here when naming it in terms.
I also think we lack a sun decay in this model..but that is also a volumetric atmospheric effect I suppose.
The atmospheric script he has ..just gives some glares around objects on the ground etc and some additional glare around the sun, but it isnīt near the type
of atmospheric look you get from vue.

Sunsky gives nice clean air sky so to speak, and even some foggier turbidity looks that is good looking, that is why I would want a mix of sunsky with the atmospherics of ogo taiki and also have ogos textured based procedural cloud layers and fog/volumetric layers for godrays and such.
the color attenuation affects hypervoxels pretty nicely and hypervoxels picks up colors and lights nicely..so it looks realistic, so in that sense it donīt seem to need much work..

however ..when the sunligh is in itīs highest peak and scatters light through the atmosphere, the global illumination also affects clouds much more, for this I fake it with hypervoxels global light illumination..but that means that light will still be as strong when you lower your sun, so in such case you need to reduce the ambient lighting because the clouds should appear darker in how much light they pick up in the inner volume in such cases, you could of course map that color to the sunsky I think in the envelopes.

A good thing about sunsky, you just apply it and change the sun or turbidity and it looks quite naturalistic, as opposed to ogo taiki where you have to mix and fiddle with to many settings to get it look right, when you do get that right it ogo taiki..it looks awesome with all the volumetrics...but slow...we just canīt have the cake and eat it at the same time it seems.
If the implementation as volumetric effect (clouds, fog and haze) is not complete, it means the atmospheric effect is not fully implemented, but it doesn't mean that a given model is not a physical sky. I can understand you don't separate these terms, but the fact is that these concepts are separated in any paper about physical skylight models. Then, in proper terms we can say we don't have a full physical atmosphere implementation, but we indeed have 2 physical skylight models.

When we use LW Fog or Ground Fog with BackDrop Color and adjust turbidy and albedo parameters in DP Sunsky, we can clearly see how they affect the volumetric fog interaction with the object. But agree, if there's no fog/haze/clouds volumetrics where the sky model can be applied, there's no way the physical sky features can affect them. Guess if LW would provide a fog&haze volumetrics (with altitude and density parameters) interacting with BDColor, physical sky models like DP Sunsky could take profit of that for the atmospheric effect since those sky models are cloudless by definition and the atmospheric anisotropic scattering and the interaction of light are calculated in that way by Lorenz-Mie and Rayleigh scattering methods.

The sharp edges issue you refer with BackDrop color (guess Volumetric AA is enabled) sounds like a LW fog issue. Don't understand well the color change issue over the object, but it sounds like if Light Attenuation needs to be increased.



Gerardo

prometheus
04-26-2014, 08:16 PM
If the implementation as volumetric effect (clouds, fog and haze) is not complete, it means the atmospheric effect is not fully implemented, but it doesn't mean that a given model is not a physical sky. I can understand you don't separate these terms, but the fact is that these concepts are separated in any paper about physical skylight models. Then, in proper terms we can say we don't have a full physical atmosphere implementation, but we indeed have 2 physical skylight models.

When we use LW Fog or Ground Fog with BackDrop Color and adjust turbidy and albedo parameters in DP Sunsky, we can clearly see how they affect the volumetric fog interaction with the object. But agree, if there's no fog/haze/clouds volumetrics where the sky model can be applied, there's no way the physical sky features can affect them. Guess if LW would provide a fog&haze volumetrics (with altitude and density parameters) interacting with BDColor, physical sky models like DP Sunsky could take profit of that for the atmospheric effect since those sky models are cloudless by definition and the atmospheric anisotropic scattering and the interaction of light are calculated in that way by Lorenz-Mie and Rayleigh scattering methods.

The sharp edges issue you refer with BackDrop color (guess Volumetric AA is enabled) sounds like a LW fog issue. Don't understand well the color change issue over the object, but it sounds like if Light Attenuation needs to be increased.



Gerardo


If the implementation as volumetric effect (clouds, fog and haze) is not complete, it means the atmospheric effect is not fully implemented, but it doesn't mean that a given model is not a physical sky. I can understand you don't separate these terms, but the fact is that these concepts are separated in any paper about physical skylight models. Then, in proper terms we can say we don't have a full physical atmosphere implementation, but we indeed have 2 physical skylight models.

When we use LW Fog or Ground Fog with BackDrop Color and adjust turbidy and albedo parameters in DP Sunsky, we can clearly see how they affect the volumetric fog interaction with the object. But agree, if there's no fog/haze/clouds volumetrics where the sky model can be applied, there's no way the physical sky features can affect them. Guess if LW would provide a fog&haze volumetrics (with altitude and density parameters) interacting with BDColor, physical sky models like DP Sunsky could take profit of that for the atmospheric effect since those sky models are cloudless by definition and the atmospheric anisotropic scattering and the interaction of light are calculated in that way by Lorenz-Mie and Rayleigh scattering methods.

The sharp edges issue you refer with BackDrop color (guess Volumetric AA is enabled) sounds like a LW fog issue. Don't understand well the color change issue over the object, but it sounds like if Light Attenuation needs to be increased.



Gerardo

I can chime in with a lot you say here..not needed to get hang up on the terms here, since I beleive we all agree itīs a physical skylight model etc.

the sharp edges has nothing to do with volumetric aa, I only use volumetric aa if I have volumetric lights in the scene, it is needed to get nice godrays, however it would be nice if the volumetric aa could be excluded on hypervoxels individually ..because a mix of volumetric lights for godray effects..that isnīt nice in terms of render speed to work with, I think that is one of the major issue when mixing volumetric lights and hypervoxels since we all know how slow hypervoxels is with volumetric aa on it, I think the actual speed render impact on volumetrics light is less than on hypervoxels.

Regarding the fog issue, i will post some images later to showcase the issues if youré not familiar with it, you could simply try and rotate the sun in sunsky if you set it to manual control and rotate it 180 degree, that will position the sun in front of the camera, and put it down in pith so t is aroun horizon level, turn on sun disc, add some mountain or hv clouds..use backdrop fog and you will have the sun disc shining in front of the objects or clouds, instead of being obscured by it, same goes with the horizon level...it is where the sunsky ends at horizon that gives birth to the sharp edge on top of objects, since the "use backdrop color" in fog setting only applies whatever color there is from the skybackdrop..and at the horizon there isnīt anything beneath it..thus it is transparent..and a high mountain pillar will in such case have fog on the top of the pillar at the level of the horizon, no matter how the camera angle is.

I donīt see any light attenuation control you are refering too, so I donīt know about that.

I got to sleep now...need to catch some sunlight in the morning, Iīll try to post samples tommorrow.



Edit...quick shots...you could of course bypass all this with rendering buffers and render out a fog pass, and mix later on for more fuller level control of it, and that is probably what most pro people do:)
but that will not let you interactivly blend clouds with the sunsky colors if you decide to use lightwave fog without backdrop color on...and that is sort of what gives the nice blend in to the sky and a nice blend on the horizon.




http://forums.newtek.com/attachment.php?attachmentid=121625&d=1398564697

http://forums.newtek.com/attachment.php?attachmentid=121626&d=1398564942
121625

121626

gerardstrada
04-26-2014, 08:56 PM
I can chime in with a lot you say here..not needed to get hang up on the terms here, since I beleive we all agree itīs a physical skylight model etc.

the sharp edges has nothing to do with volumetric aa, I only use volumetric aa if I have volumetric lights in the scene, it is needed to get nice godrays, however it would be nice if the volumetric aa could be excluded on hypervoxels individually ..because a mix of volumetric lights for godray effects..that isnīt nice in terms of render speed to work with, I think that is one of the major issue when mixing volumetric lights and hypervoxels since we all know how slow hypervoxels is with volumetric aa on it, I think the actual speed render impact on volumetrics light is less than on hypervoxels.

Regarding the fog issue, i will post some images later to showcase the issues if youré not familiar with it, you could simply try and rotate the sun in sunsky if you set it to manual control and rotate it 180 degree, that will position the sun in front of the camera, and put it down in pith so t is aroun horizon level, turn on sun disc, add some mountain or hv clouds..use backdrop fog and you will have the sun disc shining in front of the objects or clouds, instead of being obscured by it, same goes with the horizon level...it is where the sunsky ends at horizon that gives birth to the sharp edge on top of objects, since the "use backdrop color" in fog setting only applies whatever color there is from the skybackdrop..and at the horizon there isnīt anything beneath it..thus it is transparent..and a high mountain pillar will in such case have fog on the top of the pillar at the level of the horizon, no matter how the camera angle is.

I donīt see any light attenuation control you are refering too, so I donīt know about that.

I got to sleep now...need to catch some sunlight in the morning, Iīll try to post samples tommorrow.



Edit...quick shots...you could of course bypass all this with rendering buffers and render out a fog pass, and mix later on for more fuller level control of it, and that is probably what most pro people do:)
but that will not let you interactivly blend clouds with the sunsky colors if you decide to use lightwave fog without backdrop color on...and that is sort of what gives the nice blend in to the sky and a nice blend on the horizon.

The Sunsky Atmosphere has a Light Attenuation parameter, it seems you are not using it in your setup. When using physical sky I commonly don't use manual control, just Sunsky Motion for correct placement of the sun. Sorry, but can not reproduce the issue you mention here. Could you attach a simple scene, please? Maybe Denis could take a look if there's an issue.



Gerardo

prometheus
04-26-2014, 09:48 PM
The Sunsky Atmosphere has a Light Attenuation parameter, it seems you are not using it in your setup. When using physical sky I commonly don't use manual control, just Sunsky Motion for correct placement of the sun. Sorry, but can not reproduce the issue you mention here. Could you attach a simple scene, please? Maybe Denis could take a look if there's an issue.



Gerardo

Thanks for reminding about that, I have however tried that before without luck, did some initial tests now, but doesnīt seem to solve the issue, will have a closer look at it tomorrow...and if I canīt solve it..post a scene setup.

If I recall correctly, I think I have forwarded that issue to denis long long time ago, but I donīt recall how the conversation was unfolded.

I really donīt like to use the sunsky motion, I canīt be as artistic with that option, hard to place the sun where I want it, unless I need an accurate sunposition for archviz or something..I see no need for it, I think the manual control affects all other settings equally to the sunsky motion, except the sunsky motion is a different way to control the sun position accuratly regarding year/position.
donīt think it has anything to do with the issues here anyway.

Michael

gerardstrada
04-26-2014, 10:15 PM
Well, all depends on the purpose, I guess. Sunsky Motion is very useful here for simulating real sun color and intensity and replacing later over clipped-sun HDRIs taken on location. This is used later for CG integration through IBL. We have tried several sunsky models from different packages and DP Sunsky is so far the one that better match real HDR skies due to the features discussed before.

Anyway, hope you find a solution for the issue you mention :)



Gerardo

prometheus
04-27-2014, 04:17 PM
youtube search results are still wonky, suddenly my clip of this cloud animation do not appear among the clips when filtering set to upload date..itīs weird..it did some days ago, but now clips uploaded weeks ago shows..but not mine, this was an issue after youtube did some changes and it still is there:(

Tags are all there with lightwave in different versions, itīs in the description too.