Chaos Corona Forum
General Category => Gallery => Topic started by: Coronaut on 2014-03-15, 20:07:15
-
Since the dawn of man... just kidding. :D
Here is something i did in last few days, finally i got some time to do something interesting(everyone say this so i had to...)
I have made this scene(staircase to my upper floor) just for one reason, to see how real, good, etc. light simulation is in corona.
Lighting is based on IES files of existing bulbs, no portals, no fake lighting, materials and surfaces are textured as they are, so there is no fake bounces.
Render PT+HD 32PTS everything def. except Exposure comp. 2 and highlight comp. 3.5 (i have raw image that i'm planing to upload later)
Post processing in Photoshop was to add lens Flare, tighten curves and levels and small white balance adjustment and also to give lens distortion, vignetting etc. since i have used regular max cam and it sucks as it lacks bunch of stuff... Also i had to guide by FOV on camera as 17mm in any camera in max don't have slightest resemblance to real 17mm...
Also i have done PT+PT and there was no visible or significant difference(at least on my preview resolution 1200x800)
Photo was taken with canon 5d lens 17-40 f4L on 17mm and as you can see its slightly different as it has some specific distortion, no flashes, no additional lights.
Critique and all kind of stuff are more than welcome.
I would just like to thank you to developers(Steve Ballmer style) :D
I will upload other images(with time stamps, raw, more photos) until then, main dish.
Also i am planing in doing Arion version there is 2.7.6 beta available.
Cheers!
-
So the second pic is render ? Not bad !
-
Yes second one is render. Thank you.
I will upload rest of it now, just to open my eyes a bit.
-
First one is RAW. Second is White balance adjusted. I have noticed bulbs are about 400-500 kelvins higher then they actually are labeled(according to white balance)
No post processing.
Photo is sharper as it is 17mm lens while corona is FOV 64.5 and that translate in different lens size... so DOF is shallower.
-
That's pretty cool, looks very realistic! :) I think it's less the renderer than your materials and texturing (and the amount of detail).
-
Thank you!
-
Very nice image, good work realistic!
-
Thanks!
-
Wow! Cool! Looks very good! Better than foto and more realistik!
-
Well, this is sort of continuation of this topic
But after realizing that(see attachment 1) it is ridiculous for me to try and replicate tree as i did menage to do so with this stairs...
Let's say that it is inspired from this photo(attachment 2).
Photo was taken with canon 5d and sigma 50mm f/1.4@f/2.0 lens with one of the most beautiful bokeh and it's only second to Zeiss Normal 50mm f/1.4.
I did menage to fuck up with leaf stumps, and yes i was too lazy to correct that :( as i was thinking it wouldn't matter and that it would be hardly noticeable.
Tree is modeled, also leaf and i didn't use any scatter plug-in(yes i know i am crazy) but i was kinda "cmon its only 100... 200... 300... hundred leafs" and that is when i realized i should use multi-scatter. And after all this self imposed mistakes i did render something at least.
Render time was 52minutes 300passes distributive render. But i should leave it for few hours...(i know... one more mistake...)
Comments and critique is more than welcome.
-
It's really nice. Although i think it could be faster. If you have problem with DoF clarity, then you can decrease GI/AA balance value. You can also hover over the spinner to see tooltip description about why it is good to do so :)
-
I love the tree scheme! :D
It looks great but the tree model could be better. I'm not saying I could do it better, though. Great project, I'll be definitely following it.
-
Thanks!
Decrease more then 16? I find it that 16 works well for this scene, as 8 gives better(cleaner) bokeh but scene is hard on GI as everything goes trough this blinds(that are dark)+translucency(and its not that big factor in this case) i tried 32, now i will do 8 samples just to test time, but will do 800 passes.
-
I almost forgot to mention something(as i was thinking about it while i was working)
On the day i took this photo sun was slightly occluded by clouds(if you know when sun is casting rays trough thin layer of clouds) but some portion is heavily occluded by thick cloud. I knew i wouldn't be able to replicate that. As if i am to lower intensity(as i did) sun will be bleached but it wouldn't have illuminating power and vice-verse.
It would be perfect if i could make occlusion with lets say map keep intensity but being able to adjust size(as it actually effect is similar to softbox if you leave it half open)
Sun is still hitting thin layer but doesn't stop illuminating(changing intensity that much) it more changes how that photons are bouncing and creates subtle effect of soft/hard light.
It is easier to replicate this on large scale scenes as effect tends to be less noticeable(and you can use HDRI with that lighting condition).
Best example away from waiting this specific light condition is when tree leafs or something transparent(sheets, curtains(not diffuse ones) are moving on light breeze and crating changing shadows with few hard highlights of sun rays.
Well it would be nice to have this feature.
-
I almost forgot to mention something(as i was thinking about it while i was working)
On the day i took this photo sun was slightly occluded by clouds(if you know when sun is casting rays trough thin layer of clouds) but some portion is heavily occluded by thick cloud. I knew i wouldn't be able to replicate that. As if i am to lower intensity(as i did) sun will be bleached but it wouldn't have illuminating power and vice-verse.
It would be perfect if i could make occlusion with lets say map keep intensity but being able to adjust size(as it actually effect is similar to softbox if you leave it half open)
Sun is still hitting thin layer but doesn't stop illuminating(changing intensity that much) it more changes how that photons are bouncing and creates subtle effect of soft/hard light.
It is easier to replicate this on large scale scenes as effect tends to be less noticeable(and you can use HDRI with that lighting condition).
Best example away from waiting this specific light condition is when tree leafs or something transparent(sheets, curtains(not diffuse ones) are moving on light breeze and crating changing shadows with few hard highlights of sun rays.
Well it would be nice to have this feature.
There's sun size multiplier... :) It retains the power but changes how big illumination area is. It's not perfect, but it could do the trick :) And to better match the leaves to that photo. I would try feeding map that's in translucency slot through output map, and perhaps increasing output value to 1.5-2 :)
-
hi, realy nice render , i like the small tree render.
by what soft you did the modeling of the tree?
-
This is what i am talking about.
Size was changed but intensity wasn't.
Well size isn't that much changed as "semi-soft" portion is letting lets poke shit in woods and say 30% of sun rays(hard light) is missing all water, dust etc. in cloud and rest of 70%(sun rays) is diffused as photons are diffracted trough vapor(water in cloud) and turned away in some other direction. I am just saying all this is way too much over the top. Sun size does change difference of how much rays there are, but this is special situation as(refer to attachment) Hard light portion of the sun is lets say 20-25% and instead of doing 1300w or whatever lux per square meter it does 20-25% of that, and that is something i believe you can't measure but i am sure someone is simulating something like that... anyhow over the top :)
Tree is modeled in max(nothing fancy) done those small branches in Zbrush.
-
Hour and 22 minutes 800passes sample 8, dof hasn't improved dramatically, i think it would be better if i was to leave it longer. But anyhow. Next time.
...those stupid stomps on leafs... what i was thinking :D
-
If there is noise in direct lighting, rather than indirect, you can leave GI/AA balance down at 8, as it will really help DoF, even though it may not seem so, and crank up Light Samples multiplier instead. That's what mostly takes care about noise from direct lighting. Environment light is considered to be direct light too, by the way ;)
-
This will be last contribution to this topic.
Here is one on first look very simple scene... but in fact it is very demanding as it have 3 sheets of glass, DOF, bunch of glossy reflection, dome(HDRI) lights, and it's very very low light.
Original photo was taken with canon5d, sigma 1.4@1.4 Iso 1600 shutter 1/100, it was sunset with intricate lighting(dark clouds with orange highlights(very bright spots) and unfortunately i wasn't able to reproduce similar highlights on chrome and aluminum parts as i didn't have that kind of HDRI...
Anyhow... I didn't like how dof cuts out it seems to me like depth falloff doesn't have right curve... Glass transparency is pretty much as it should be and it doesn't kill high lights like in Vray.
p.s. what test would this be if i wouldnt be able to fuck up... this aluminum rail is different as i was too lazy to measure it on the window i did take picture instead i did on the one above my head and it has slightly different one...
-
Vray versions. As i did mention in vray versions there is clear absence in highlights inside that metal frame of the glass... Also i had to remove glass out of GI as it was too dark... and cuz i want to replicate real world condition not fake(vray always had this problem with glass absorbing too much rays)
I didn't have this absorption problem with vrayRT but i did have(unsupported material and all that scheisse...)
Render times
Corona 8:30(hours) 422passes i think its too long as it was done on 28threads in DR mode but i didn't check thin glass(on purpose) 8pt samples...
Vray 3:20(hours) Pretty high settings(except leaving glass out of GI solution) i left caustics and all that stuff.
Vray_RT 30minutes(resolution almost 50% lower) Two GTX780 cards
Arion(i should spend that money on drugs when i was in Amsterdam few months ago)... Disappointment, i am sorry for lost time when i was converting scene... It just couldn't done anything...
I will try 100 passes with corona just to check if i can get faster times.
That is about it. From this different scenarios.
Cheers people!
-
I think that
refraction transparency of your glass material is to low. Look how big difference in sky's exposure betwen direct and through glass views if you compare it with photo.
-
Yes, that is what happens when refraction is on, and when you have 2 completely different sky's and 3 layers of glass. But i am sure you didn't wanted to state obvious, so please, tell me if you have any trick down your sleeve, how should i increase transparency of already unrealistically transparent glass?
Also if you think that this corona version has major discrepancy between glass and air, check out some other rendering systems, and if there isn't any i would love to see some of this kinda tests from you.
Also to take into account how "real" lens doesn't have ideal Optics(why do i write with what lens photo was taken?) but instead has vignetting and bunch of different kind of distortions, not to mention that i have filter mounted on lens... Maybe what you see as brighter sky on render and same brightness on photo is in fact what's wrong. Or maybe something completely different. I get how this is gallery section but in fact this should not be in gallery as maybe there is some people who will look for something that you tend to overlook as it is more subtle than glowing green or blue.
Anyhow, as there is "Thin" option there just so you can get airy, unobstructed GI, why not click it, right?
And as i don't want to be jerk and to be "talking shit and stuff" or give your comment any deeper meaning here is what i am talking about.
http://www.roofingsuperstore.co.uk/browse/pitched-roof-windows/velux-replacement-panes/laminated-8.html 0.78 light transmittance...
http://en.wikipedia.org/wiki/Transmittance <---more on that theme
If you already thinking, please continue to do so.
Cheers.
-
In Vray, GI is what takes care of caustics, so if you have left the glass out of GI solution, then the comparison is very unfair. So either enable GI for glass, or in Corona, disable Caustics on your glass material ;)
-
Yes, my previous post and render above your comment is same "solution" as in Vray but it is in corona. As you can see it is 3 time faster then Vray.
Anyhow it would be faster even with GI caustic as 100 passes would be done in less then 2 hours.