I have been trying to understand how the Vray Sun/Sky + Vray Physical Camera combination relates to real world conditions. To test, I created a simple physical model that I photographed and built a similar 3D model that I rendered trying to match conditions and settings. I’ve shown the results below.
I immediately noticed that the Vray Sky seems to have much more cyan than does our sky in the mountain west US. No matter how I tried to adjust the color balance, I couldn’t match the real world. I also tried putting the sky in an output map and adjust but that didn’t help (but I don’t know that much about how to use the output map).
The second thing I noticed was that the light bouncing off the cube was more intense in the model than in the photo. I played with the Primary and Secondary bounces but couldn’t match the photo. I didn’t try reducing the “generate GI” values, which would probably help.
I also discovered that almost without exception, the Vray Physical Cam was one stop brighter than the real camera. I either had to reduce the f-stop one stop or the ISO by 1/2 or increase shutter by 2.
Overall, I was very impressed at how quickly I was able to approximate the real world by using many of the default settings, the exception being the white balance. I hope that anyone who has had good luck replicating the real world will offer suggestions that will help me understand how to get better results.
Craig
Settings:
Gamma (display, input & output) 2.2
Mapping Linear multiply, 1,1,1
GI Primary IR, Secondary Brute Force
Medium Preset, HSph50, Int Smpls 20
Overcast Day Physical Cam WB RGB correction: 88, 157, 238
Sunny Day Physical Cam WB RGB correction: 242, 255, 250
Rendered to Max VFB, saved as 16 bit TIFF at 2.2. No PhotoShop corrections except to add text and save as jpeg.
Other settings shown in the images.
I think the difference in the f-stop and color bleeding comes from the materials. Reducing the brightess of the materials accordingly will give you better results both with respect to the bleeding and the f-stop.
excellent work. Yes, as vlado mentioned, there was another post similiar to yours where the poster couldn’t get the exact same values, and it appeared his materials were too bright. Yours are pretty darn close, as your images show.
yes pretty close…good test too. But, I would say that the sky color depends also on world positioning, day of month, year etc.
The tests are pretty close too!..
Short answer, you will have to adjust material intensities down, even if you do use phys cam exposure. Because white 255,255,255 is not the color of a white object, it is the maximum brightness of the image.
I don’t know the math, but there’s a way to sample a swatch in your color card, compare it against the swatch in the rendered example, and use the difference as the white balance value in the vray phys cam.
also with the reference photos you shot having the camera set to auto white balance can mess things up if your trying to reproduce it.
You need to fix the white balance on the camera then match it in the vray camera.
I added some month ago a wish to the wishlist and still think it’s a valid and usefull wish: to apply a colorprofile of defined colorspace or different real rollfilms to the renderoutput or embed colorprofiles to the renderoutput to enable working in photoshop in a defined colorspace.
I don’t know how to achieve a correct overcast illumination. Is it correct to set size multiplier to 1000 to get an overcast illumination?
I’ve done a some quick tests and seems that setting sun size multiplier beyond a certain value (213?) suddenly get the same results than disabling the Vray sun … seems weird to me.
Thanks all for the replies. I will re-shoot the photos using a different white balance, perhaps doing a real white balance on the white material? I will also try reducing the material’s brightness as well as following the other suggestions posted here and repost when tests are complete.
Interesting comparison. One thing I was thinking after talking to a painter recently is that maybe the "white"card isnt actually 255/255/255. In the photo’a the white looks to possibly have a bit of grey in it which isnt noticable so much in the areas where the sun directly hits the card, but it effects the shadowed areas. The painter told me that they add a few drops of black into white paint…apparently makes it a longer lasting finish or something, so in addition to what everyone else said, the card may actually be something like 250/250/250. Everything i’ve said could be a complete load of crap though!
You are right. I originally used a value of 240 240 240 for the white cardboard but after Vlado’s comments, I lowered it to 225 225 225 and it is rendering much better. I also compared the original white board I used to build the physical model to a very white paint chip and the board is quite grey, about the same difference as 225x3 vs 255x3 I see on the monitor. I think I’m getting close.
(This post is not an offence against the author.. it´s just my general
opinion about physical correct rendering.. once again 8) )
I still don´t get it why people are still making their life so hard by trying
to do everything physicaly correct. This is at least the 5th post I read were people are having problems to recreate their photos with physically light models. It depends on so many factors.
Each camera has a different color curve. Different white balance. The sky itself can differ very much from day to day.. and from the part of the world you are currently located. You would need to measure every single material.. to make it physically correct. Let´s face it.. a physically correct rendering simply doesn´t exist… the only physically correct thing that makes sense is a physically correct light distribution.
It´s completely pointless. For example… if you look at some image you rendered and for some reason you don´t like the sky..
let´s say.. it´s to dark. Do you think for yourself.. “it´s to dark”.. or.. “there is to much ozone in it”.
We as 3d Artist don´t have all these limitations photographers have .
So why should we make this huge step backwards and bind ourselfs to all these limitations that just exist in real world. Why should I adjust my exposure in something abstract thing like F-Stops if I can have gain/gamma/saturation. Why should I mess around with ozone and turbidity if could just use a gradient were I can adjust every color as I like it. Are we artist or physicist ??
Well.. and I´m sorry.. if someone is not able to make a sky with a simple gradient… he won´t come to far with ozone/tubidity.. and the other cool words.
upss.. did I really just say all of these things ?
:lol:
I think you miss the point entirely. I’m not trying to make physically correct renderings, just visually correct ones. Nearly all my exterior work is 3D / photo composits and my goal is to try to make them look seamless and real. I’ve used many different methods over the years to obtain my goal. I’m just trying to see if there is a better tool available for me to get there and posting my findings so others don’t have to recreate the tests. I value the constructive comments given and I’m sure others benefit as well. Who benefits from your redundant spoutings? (two more cool words)
In my continuing tests, I haven’t had too much success with the Vray Sky. I just don’t have enough control over its settings (or just don’t know how to use it, which is more likely). Until I get more familiar with it, I’ve decided to show a well used approach just for comparison. Since the Vray sky is simply an educated guess at a typical sky we can guess as well. For this rendering I added a plane representing the clouds and sky and added a light blue VrayLightMTl set to 40. I had to crank up the blue to compensate for the Turbidity of the sun which can’t be set lower than 2. Where I work, in the US Western Mountains at 4500 feet, there isn’t much Turbidity at all so it would be nice if that number could go lower. Also, I tried several times to get the shadow of the cube to be as light as the photo, but each time the whites blew out. Darkening the white texture didn’t help. Comments and suggestions would be great. (Note that I turned the model 90 degrees to avoid sun-glare on the lens and sheen on the foamcore.)
Photo - Taken at 12:00 noon on a 50% cloudy day, camera WB set to daylight. Shutter 1/500 f10 ISO 200
Render - PhysicalCam shutter 1/500 f10 ISO 200. WB set to Daylight. Light Turbidity 2, Intensity Mult 1.0, Size Mult. 5. Sky plane to represent clouds: VrayLightMtl at 40, color at R187 G235 B255.
