How to get better VRay renders (aka working in linear space)

[flipside]

Hi throb,

I looks interesting, but I don't get I think.

Why would you do all this adjustments, then render through backburner and it's all lost?

I also don't see the difference between adjust curves in the VFB or just save the image and do it in photoshop or something. (except that you don't need PS this way)

All my tests are very washed out with the curve you use in your tutorial.

Please try to clarify this for a 'curves-newbie'

wouter

[lukx]

I didn't do any test but I guess that the same effect you can achieve with exponential color mapping.

[throb]

Why would you do all this adjustments, then render through backburner and it's all lost?

It's not lost. Your compositing app (nuke, shake, fusion, combustion, etc) all have viewing LUTs that will apply the same curve when you view that output

I also don't see the difference between adjust curves in the VFB or just save the image and do it in photoshop or something. (except that you don't need PS this way)

The reason is that when you have your image prepared using this viewing lut, you have an image that is stored in linear floating point. That means when you mult data, it's doing a correct mult and not a non-linear mult. If you save the image with the curves baked in, then you will get non-linear color corrections on your image.

All my tests are very washed out with the curve you use in your tutorial.

Did you load a fresh scene or load one you already had done before.
If you loaded a previously created scene, you are going to get bad results because all your mats and colors are set for non-linear workflow versus a linear workflow.

To answer the other question:

I didn't do any test but I guess that the same effect you can achieve with exponential color mapping.

As I explained above...if you bake in this correction then you negate the idea of working in linear color space. Why? Well, you create an image that is -in the image- corrected for your monitor's fault. This means that you are now working in a non-linear space.

Hopefully that makes more sense.

rob

[Trapezium]

Ahhh, I see now. Then from what you've just explained, no, the Studio RGB has nothing to do with it. I didn't realize hat this workflow was to be saved as 16 bit plus and managed in the comp software. You should add a little section on saving and viewing the rendered images in an appropriate compositing software.

BTW, I can't see the third image down "http://throb.net/media/linear/data_in.png" anymore

Thanks again,

--Jon

Image is fine now

[throb]

Ahhh, I see now. Then from what you've just explained, no, the Studio RGB has nothing to do with it. I didn't realize hat this workflow was to be saved as 16 bit plus and managed in the comp software. You should add a little section on saving and viewing the rendered images in an appropriate compositing software.

You are 100% correct. I have added new images that help tell the story of what happens to your data from creation to output to correction of output. Go back to the URL and you'll see what I mean.

Unfortunately I only have Nuke to composite with here so I can't show you guys any settings in comp packages that will allow you to account for sRGB correction.

BTW, I can't see the third image down "http://throb.net/media/linear/data_in.png" anymore

Oops. You caught me in the middle of the edit. All fixed now

enjoy

rob

[cpnichols]

hmmm... Rob let me try to see if I can explain this for those that may not get it.

If you "light" a scene accuratly without looking at it through a LUT (the correction curve throb is talking about)... Your darks tend to be very dark. The reason is that your image "viewer" is looking at a floating point image in Linear Space, and not Monitor Space.

So your solution is to add more lights, increase the amount of GI, etc... As you know, this may actually increase your rendering time. Now, if you continue to work in "linear space" then you are fine. Which is what most of you do.

However, your monitor actually exists in sRGB space, you can compensate for that be doing a reverse of the Gamma curve of your monitor. You are now working in sRGB space. Photoshop, Nuke, Shake, etc... all have the ability to work in sRGB space. Throbs correction curve is simply giving the Vray VFB the same ability. The added bonus is that you don't have to pump up the darks through lighting anymore. However, if you have a scene where the lights have already been pumped up, then it will look washes out as flipside has noted.

To give you another analogy. You know those Bose "noise cancelation" headphones? They remove the ambient noise be ceating a reverse sound wave. That is basically what the monitor curve is doing. By simply increasing the amount of light to compensate for the darkness, it is like tunrning up the volume. Ok... don't know if that is a good analogy.

BTW... Throb really knows what he is talking about... trust me... I work with him every day.

[throb]

To give you another analogy. You know those Bose "noise cancelation" headphones? They remove the ambient noise be ceating a reverse sound wave. That is basically what the monitor curve is doing. By simply increasing the amount of light to compensate for the darkness, it is like tunrning up the volume. Ok... don't know if that is a good analogy.

this is a great analogy!!!

nice!

rob

[Da_elf]

ok. so let me get this straight. we do a render. it looks dark. this is because our monitors are screwing it up for us. the image really isnt as dark as the monitor makes it out to be. soo. we use an inversed curve to compensate for the monitor and we make it brighter. brilliant. great stuff. now then. what happens when you print it? we have made it brighter to be able to see it properly on our monitor which was making it darker. will that mean the image wil print brighter than what we see on the monitor and thus making the image appear too bright?

[throb]

ok. so let me get this straight. we do a render. it looks dark. this is because our monitors are screwing it up for us. the image really isnt as dark as the monitor makes it out to be. soo. we use an inversed curve to compensate for the monitor and we make it brighter. brilliant. great stuff. now then. what happens when you print it? we have made it brighter to be able to see it properly on our monitor which was making it darker. will that mean the image wil print brighter than what we see on the monitor and thus making the image appear too bright?

on that I am not sure....

do some tests maybe?

i deal with film and video and so this isn't an issue.
I know that if i did a proof and it came back dark/light that i would have to correct for it.

[davision]

Well, if I may add to this confusion... Da Elf, you are sort of on the right track. What I believe throb is saying, is... you only PREVIEW the image on the monitor (with the correction curve -- i.e. gamma 2.2) to make sure everything looks okay. You don't save it that way. When printing (in theory) you would use the uncorrected image, since printing is supposed to have a gamma of 1.0 (or linear space) where the darks will print more brightly naturally. So, to answer your question, if you correct it, then save, then print... it will be too bright.

Before, though, I said "in theory" because my experience is that printing uncorrected images end up somewhat dim. They still do need some "correction" just not as much as a monitor does. But this could be opening a can of worms...

Also, many of us don't ever print our images -- they end up on computer or a TV so then we need that highly corrected image, but the idea is that you'd save the uncorrected "floating point" file and do the correction in your post app.

My questions are...

Throb, what do you set the input and output gammas to in the preferences... leave at 1.0???

Also, why do the correction in the Vray VFB, if the regular VFB does it for you (if you set the display gamma to 2.2)?


Also, what is a good 'floating point' file format to use for saving renders?

I'm used to working with 24 bit, so I don't know 48-bit or whatever it's called.

[cpnichols]

ok. so let me get this straight. we do a render. it looks dark. this is because our monitors are screwing it up for us. the image really isnt as dark as the monitor makes it out to be. soo. we use an inversed curve to compensate for the monitor and we make it brighter. brilliant. great stuff. now then. what happens when you print it? we have made it brighter to be able to see it properly on our monitor which was making it darker. will that mean the image wil print brighter than what we see on the monitor and thus making the image appear too bright?

Actually there is an easy answer to this. When you open it in Photoshop, make sure you are using an sRGB lookup. Remember all those wierd color options that keep popping up in photoshop that you don't really undertand what they mean? They should make a little more sense now. THEN, when you print, make sure it sends it to the printer with an sRGB curve to it. Photoshop has the ability to do that.

and Davision... I already know that answer that throb will give you... EXR. and I could not agree with him more. Of course, he and I have the luxury of working in Nuke which is VERY EXR friendly.

[davision]

cp... wouldn't that method you describe for Photoshop APPLY the SRGB curve to the image when printing --- and therefore apply a monitor compensation curve to a printed image??? That doesn't seem right... you would want to strip the curve out and print more linearly, no?


Oh, and regarding EXR -- I don't think AfterEffects supports that though. Any alternatives?

[Trapezium]

RPF will save the information necessary. Combustion handles them natively and I believe the latest AF will import them also.

--Jon

[davision]

Okay thanks... I thought rpf might be the route to take...

[Trapezium]

no problem.

Not sure if AF has an LUT viewer though...

Not too sure about what's going on with the one in Combustion yet either

--Jon