Universal V-Ray settings (well, almost)

[cpnichols]

Thanks chris, This is what i have been doing, with great results, and with little or no p.s., I think i am using methord "B" by doing a gammar correction for my color mapping, using DM=1 and BM 0.4343 i do not apply the curve in viewer and the image looks the same in PS as it does in the viewer. I am using the latest offical Vray build 1.47.03
this is why i am confuse when vlado said you will get incorrect results.

Cecil G

Actually that is not true... if you save your image as a tiff (8 bit clamped) with the color mapping option. It will look the same. HOWEVER, if save your image as a floating point unclamped image (HDR or EXR) PS will use the monitor curve and apply it to your image... WITHOUT (and this is the key) altering the data. It applies the change (automatically) in the viewing of the image, just like you have done in the Vray VFB.

Using the color mapping is fine... it is a crutch... but eventually you will see the advantage of working in full float and you will no longer want to work in 8 bit color space... at which point you will have to stop using the color mapping.

[priad]

great settings Vlado and great ring example!!

[vlado]

The rings are from Thomas An's challenge over at CGTalk, I thought they would make a good example for the settings.

Best regards,
Vlado

[throb]

The rings are from Thomas An's challenge over at CGTalk, I thought they would make a good example for the settings.

ok, so ring 4 is 1/4 the render time of ring 3.
what's the diff in the settings?

[vlado]

ok, so ring 4 is 1/4 the render time of ring 3.
what's the diff in the settings?

The only difference is that the slower image had path-traced GI caustics, while in the second one the caustics are done with photon mapping. The caustics turned out with wrong intensity in the photons image though, because of a bug in the build I was using, so I'm re-rendering this now.

Best regards,
Vlado

[Sawyer]

So Chris um I don't know what the benefits are with a floating point output. I know the value of using hdr's in an image and I could imagine there is a value having that same amount of information in an illustration but I'm not there yet. I don't know how I would use a floating bit image.
I imagine I would have greater control of the exposure.


Thanks

[cecil gittens]

Chris,
I do not have a clue about floatingpoint images, so could you explain ferther. or post an example.
Cecil

[joconnell]

Hi Folks.

Floating point images are mainly of use in compositing and visual effects where you are doing a lot of colour correction to try and match one thing to another.

Regular images are stored with a colour range between 0.0 (black) and 1.0 (white) with all of the shades of grey and colours in between. you may also be used to seeing this as 0 - 255. The brightest colour that is possible to save in this type of image is pure white which will have a value of 1.0 or 255 depending on which program you are using.

Floating point images allow you to store values and colours that are far brighter than 1.0 - your image will always display them as pure white but it contains information that may be bright than 1.0. Take for example if you have a vray light in your viewport with a colour of white and a multiplier of 30 - when you render the image with the light in it, it will look pure white but if you save the image to a floating point image, the actual colour value of the pixels for the light will have a value of 30.0 - thirty times brighter than pure white.

You may think what's the point in using a floating point image if you can't actually see the difference but it's a huge benefit when it comes to colour correction. Say for example if you have a really bright highlight on an object or perhaps a reflection of a bright light source, lets think of something where the colour value of the highlight / reflection is around 1.5 - in otherwards one and a half times brighter than pure white. If we render this to a non float format, the highlight will get "clamped" - again if we remember that in a non float image, the brightest image colour is 1.0 or pure white - clamping or clipping a colour means that it's value is limited to the range of the image so our 1.5 highlight will be cut off at 1.0 - it will look pretty much the same when we view the render but it will have problems in colour correction later on.

Say for example we take this render into a compositing program and apply a brightness and contrast command and take the brightness down to 50% - again there's few cases where you'd do this but just go along with it. 50% brightness of our 1.5 specular highlight should be 0.75 and if you use unclamped colour and a floating point format it will be. Our clamped or non float image will have cut off the 1.5 value at 1.0 and so after the colour correction to 50% brightness, you'll get a highlight with a value of 0.5 which is totally incorrect. Float basically keeps all of the excess data intact.

Another benefit of using a float system is when you stack one colour correction on top of another. Say for some stupid reason you have an image and you apply a brightness operator and brighten the image to 500% of it's original values. This will probably mean that most of the image will be completely blown out and white. Now say you put on a second brightness operator and set the brightness to 20% - 500% x 20% = 100% so in effect the two operators cancel each other out. If you are using a non float compositor, the first brightness operator will have made most of the image white and cut off the values at a value of 1.0 . When we apply the second colour correct to bring the image back down to 20% brightness and reverse the process, our image will be totally different - because the white pixels got cut off at 1.0, they'll get colour corrected back down to 0.20 so we're not getting back what we put in.

If we were using a float system, it allows us to store colour values way above 1.0 so if we do the same thing, a pixel with a value of 1.0 or full white can be made 500% brighter quite easily and now have a value of 5.0 - you wont see the diffference but the data is there. Now when you use the second brightness operator of 20%, 20% of 5.0 is 1.0 so you're getting back the exact same image with no data loss.

Basically the floating point image method gives us far higher quality and means we can treat our images more harshly with less problems.

[flipside]

Nice explanation!

What format do you recommend to save your images in?

[rpc212]

Open EXR is the preferrable one I believe. Of course as long as your compositor/color correction app can open Open EXR.

[joconnell]

Nice explanation!

What format do you recommend to save your images in?

I reckon EXR is a good bet but being brutally honest the place I work in doesnt take heed of a lot of this stuff so I never use float myself - I'm always doing tv work and most of the compositing stuff is done using flame which doesnt support float. When I get an upgrade to after effects 7 I'm gonna look at this again though.

[Frances]

If my client and I agree that our screen and printed output is correct (enough), do I need to obsess about why I just don't get the linear color mapping thing? In addition to that, I don't see much point in worrying over it until I can afford a hardware calibration device, right?

I fully understand why this issue is crucial in the film and VFX industries.

[cpnichols]

If my client and I agree that our screen and printed output is correct (enough), do I need to obsess about why I just don't get the linear color mapping thing? In addition to that, I don't see much point in worrying over it until I can afford a hardware calibration device, right?

I fully understand why this issue is crucial in the film and VFX industries.

Well if you DO use it, it will change the way you light. Also, based on what I can tell about the way that Vray is going with more real world lighting situations, it seems that is will be critical to the near future updates of Vray.

Keep in mind that, you can get at a similar lighting as you do now. You will simply get there more "correctly." For example, you will use less fill than you do now. Also, if you want more contrast in your image, nothing is stopping you from adding the contrast in post. When you work in full float, your image can scretch like bubblegum.... not so much in 8bit.

[sv]

Say for example we take this render into a compositing program and apply a brightness and contrast command and take the brightness down to 50% - again there's few cases where you'd do this but just go along with it. 50% brightness of our 1.5 specular highlight should be 0.75 and if you use unclamped colour and a floating point format it will be. Our clamped or non float image will have cut off the 1.5 value at 1.0 and so after the colour correction to 50% brightness, you'll get a highlight with a value of 0.5 which is totally incorrect.

Testing this out using a 32-bit .rpf image in comparison to an 8-bit jpg image in combustion, i can see the colour values changing as you have described, - however, if i load the images into combustion and pre-multiply the image, this seems to clamp the colours and the colour values are wrong. What is the effect of pre-multiplying the image - this is something that i usuall have to do when compositing images, otherwise you get a horrible black line around your layer?

I appreciate this is a little off topic, and really more of a compositing question, but can anyone give an explanation of this?

[flipside]

(*) Set the QMC Sampler "Adaptive amount" to 1.0.

With these universal render settings, if you are using ir map for first bounce, would you still set adaptive amount in qmc sampler rollout to 1.0?

Why did you do that in the first place?

Also, why do I get much less noise in vraylights area shadows when using qmc AA 1/4 compared to 1/100? My lights subdivs were both at 25.

thanks,

wouter