Questions about the future of Vray sun/sky

[tmbarker531]

Personally, I don't particularly love using HDRIs for a variety of reasons. I've become kind of obsessed with lighting that's as technically realistic as possible without the use of images someone has taken from real life.
I thought the hosek model was a big improvement when it was released, and when renders are saved out as 32bit exrs it is very nicely controllable in post work. But I've started to wonder more about how it works.

The biggest confusion I have is how the system works after the sun is pulled below the horizon- it seems to change realistically all the way through sunset but once it's pulled below the horizon the sky will stay very orange/yellow. Maybe it's just me, but it seems like in reality this is not happening and instead the sky is left much bluer once the sun actually has set? I've gotten around this dusk problem pretty well by positioning the sun somewhat low in the horizon and turning it off but keeping the skylight on, but clearly this is faking a solution.

In relation to that question I was wondering if someone could point me to more information on what the preetham, cie, and hosek models are based off of and what are the known limitations of these models?

Lastly I was wondering if anyone knew of someone who has tried rendering clouds that are realistically created using some kind of particle system? Or some kind of tutorial on how it could be done. Cloud control seems like the next large territory for me to try to experiment with and it is the main factor in having to occasionally still use HDRIs.

Thanks for all future help, also thanks to everyone at chaos group for what you do, you're my favorite people

[mohammadreza_mohseni]

The biggest confusion I have is how the system works after the sun is pulled below the horizon- it seems to change realistically all the way through sunset but once it's pulled below the horizon the sky will stay very orange/yellow. Maybe it's just me, but it seems like in reality this is not happening and instead the sky is left much bluer once the sun actually has set? I've gotten around this dusk problem pretty well by positioning the sun somewhat low in the horizon and turning it off but keeping the skylight on, but clearly this is faking a solution.

it does happen it reality as well.

In relation to that question I was wondering if someone could point me to more information on what the preetham, cie, and hosek models are based off of and what are the known limitations of these models?

I can offer you read some great articles out there about these models. I put links of some article that I have read myself in this topic which may help you.

http://climate-based-daylighting.com...evic_chap5.pdf
http://cgg.mff.cuni.cz/projects/Skyl...int_lowres.pdf
http://timothykol.com/pub/sky.pdf
https://people.mpi-inf.mpg.de/~jgrya...etering/LM.pdf
http://www.graphics.cornell.edu/Kide...ture(2014).pdf

[Neilg]

^ did you read any of those before posting them? CIE and hoesk are daytime only and that's mentioned in the first 3 of those I clicked at random.

The sun/sky model currently doesnt support any below the horizon atmospheric effects - you cant do dusk with it. need to go back to hdri's for that.
Vlado said it's somewhere on the to-do list to add more models - some of which do support dusk, I just dug up an old post.

[mohammadreza_mohseni]

^ did you read any of those before posting them? CIE and hoesk are daytime only and that's mentioned in the first 3 of those I clicked at random.

yes of course. Hosek is daytime and we are talking about daytime, aren't we?
night time start after astronomical dusk time.

[tmbarker531]

yes of course. Hosek is daytime and we are talking about daytime, aren't we?

Well no... my question was purely about what happens when the sun is pulled below the horizon...specifically in all the stages of dusk shown in the diagram you posted.

Good to know that it's on the to-do list Neilg! And thanks for the reading links Mohammadreza

Still would love to know more about realistic clouds made from particle systems...

[tmbarker531]

ah so you're saying it's accurate through dusk, all the way up to 18 degrees below the horizon? still not sure if i fully believe that

[Neilg]

night time start after astronomical dusk time.

[ATTACH=CONFIG]36559[/ATTACH]
[ATTACH=CONFIG]36560[/ATTACH]
[ATTACH=CONFIG]36561[/ATTACH]
[ATTACH=CONFIG]36562[/ATTACH]
[ATTACH=CONFIG]36563[/ATTACH]

no - daytime is before sunset and dusk. Dusk is not supported & even sunset is a bit shit because it misses what happens in the opposite direction to the sun. your images kind of prove that... the second the sun goes below the horizon the result breaks.


Dusk looks like this - https://i.ytimg.com/vi/2st91huy5Os/maxresdefault.jpg
Instead of the blue coming up as the sun nears the horizon and light scatters through the atmosphere miles above us, nothing happens. None of these models simulate that upper atmosphere light scattering.

[tmbarker531]

even sunset is a bit shit. your images kind of prove that...

haha, yeah that burnt orange is always pretty nasty and feels unrealistic. but in the jpg you just linked- it seems like so much of the look of that image comes from the atmosphere which seems sooo much more complex than just enabling aerial perspective or using environment fog

[mohammadreza_mohseni]

all sky models are inaccurate in some points. they all have some limitation just like Hosek when the sun drop to 10 degrees start to show some inaccuracy. all I was saying that in dusk sky suppose to have some colors as you obviously know so this not strange. but it will be not physically correct whith hosek sky model as Neilg mentioned before.