Am I switching this on or off for true LWF? All my images lately seems to be stuck at the same quality and it feels like I’m not progressing as an artist. I usually have problems with burnt out areas.
If I switch this OFF dont affect colors and use reinhard, the burns goes away. But obviosly when I switch ON dont affect colors the burns comes back as reinhard then has no effect
Reinhard is not Linear at all, hence it is not a linear workflow. For true linear workflow you have Linear Gamma Correction, Gamma 2.2, dont affect colors on. Then you are saving out a true linear image. Colormapping and/or grading is to be done in post then in an app that supports working in linear.
Regards,
Thorsten
you check don’t affect colours when you don’t want any gamma correction to be be applied on your output, vray will still sample the image correctly though. it makes sense if you work linearly and save to a linear format, so that you can always check how the output will look toggling the srgb button in the vray fb, or if you do your colour mapping in post.
Can I do color mapping in post? (like a reinhard type) in Photoshop? if not, then how about in Composite?
Well Reinhard is a Blend between Linear and Exponential Mapping controlled by the burn value. Should be doable, but wouldnt know exact how-to as using neither Reinhard nor PS here he.
Regards,
Thorsten
What color mapping you use? I know you guys are using Nuke, but what color mapping? I know you render linear, so yes the question is what colour mapping you guys apply in post?
Well none really. We render linear, save linear to exr, load linear to nuke and work linear. The only “fixed” colourspace conversion that happens is to sRGB for final output. The whole look etc is defined via standard grading etc, no need for specific colourmapping operations.
Regards,
Thorsten
so then do you deal with the common “burnt out areas” when grading? Or you dont get any burnt out issues in the 1st place? Also, does your images come out “greyish” or are the colours pretty much perfect straight from render?
I think that thorsten is speaking of something at a much higher level than you understand. Linear workflow is a workflow, you don’t really get the advantage of working like that until you start using something like Nuke. I think that there are very very few people that actually understand how this work, and utilize it properly. Most of us, we’re just using the gamma hack to change the behaviour of light in our renderings, and to be honest it works for me.
The whole idea behind linear workflow is that by applying stuff to a linear image, it will look different when seen in sRGB. For example, if you apply gaussean blur to an image in srgb, it will look different from an image that had the blur applied while in linear, and then brought to srgb. Correct me if im wrong.
*edit*
I forgot, is working with floating colors a must in the so called “linear workflow” or you can have a clamped image? I think that the burn out areas problem is a very subjective one. Reduce the exposure and you won’t have burnt out areas.
Edit: Erm…that turned out a “tad” longer than i thought. Sorry for that. Dont be put off by the formula in the middle…no need for math, just added it to illustrate the complexity of sRGB 
Hey there,
just to put a few things straight. There are a lot of ppl that get along without LWF just fine. And if you yield the results you want/need there’s absolutely nothing wrong with that. There are quite a few very different reasons for LWF.
To answer your question pailhead, float is a must for LWF, but NOT because of clamping or values above 1. A floating image can also be clamped to 0-1 and STILL be perfectly Linear and STILL be way better than a non floating image. Keep in mind that the BIGGEST advantage of float images is not values above 1. I consider that as a nice addon. The real advantage is higher color precision (you get more values between 0-1 than between 1-Infinity with floating point math btw hehe).
The reason float is a must for LWF is indeed this higher color precision as you would be lossing a lot of color precision on the PERCEPTION side of things in linear colour space. The whole idea of sRGB is to have better colour precision in the brightness range that is best percieved by the human eye without needing more space (both memory and HD wise). With floating point this is not only not needed, but also introduces quite some issues.
One of the reasons for LWF in the first place (during rendering) is different distribution of Light (for example the often observed strong colour bleeding and similar effects). While this is perfectly fine if you are either just starting with 3D or after getting used to it, it mostly puts of 3d artists in the first place because they are very used to working in sRGB space and their workflows/“default settings” all of a sudden dont yield the result they are aiming for. One thing ppl seem to have a very hard time with it that they have “too few blacks”. So if you choose a dark grey it is brighter with linear workflow…and that standard colour value for black plastic all of a sudden doesnt work at all anymore. Having float precision you can rest assured tho that there ARE enough blacks
Moving on to the post (be it PS, nuke or whatever tool you use) sRGB poses different issues (and hence different reasons to use LWF).
There are quite some cases (facing them everyday in fact) where calculating in sRGB simply creates WRONG results. This may not be obvious when doing tweaks to a finished beauty rendering but they pop up everywhere if you do more complicated post like rendering in layers, seperating elements etc. Can you say “fringing” ?
Generally linear is actually the EASIER and more logical approach. sRGB is a pretty crazy hack if you think about it. The idea of a fully linear pipeline is simply to keep your images “as is” through all stages.
Just read up on the Gamma curve that sRGB really is. It is not a simple gamma 2.2 in reality. Put simple it is a linear segment between 0 and 0.04045 and a gamma 2.2 segment between 0.04045. What kinda wacky unpredictable hack is that ? (The reason is to prevent the curve from becoming perfectly horizontal at 0 btw).
Here’s the corresponding function:
v < .04045 ? v / 12.92 : pow((v+.055)/1.055, 2.4)
Not exactly straight-forward is it ?
I just wrote this to explain that Linear is actually easy. 1 is 1, 0 is 0 and 0.5 is 0.5 period.
Now the problem is that all monitors have their sRGB correction built in and you have to feed them sRGB luminances in order to see what you are really doing. This is what the sRGB button in the VRay VFB does and what an sRGB Viewer LUT does in nuke.
So on the practical side this is what LWF boils down to
- Input images like textures either need to be in linear space or corrected on load
- Colours need to be set in linear space
- Rendering is performed in linear space
- The image is PREVIEWED using an sRGB viewer correction but SAVED linear and in floating point
- Post production is PREVIEWED in sRGB again, but all the math (be it grading or transformations) happens in pure linear floating point.
- Final output is determined by the format and target. So you’d output to sRGB for jpg, QT, AVI etc. but might stay linear for your internal final rendering.
That’s pretty much it. Sorry for the long post, and if i fubared things somewhere feel free to correct me!
Regards,
Thorsten
I get all that, I don’t however have an answer to my original question
In other words my renders straight out of render doesn’t look right. Are they supose to look right?
That depends on your definition of “right”. You can of course fight burned out areas in post. With float you can easily adjust exposure or non-linearly tweak the luminance as needed. This needs float tho usually.
Regards,
Thorsten
What is the output gamma on the .exr format when saving a linear image? I get a bit confused with this, and i don’t have my max now to check. I think i set it to 1.0. I think i have to override to 2.2 when saving 8bit images to get them in sRGB. But when saving .exr at 1.0 they are also in srgb, not linear, i think 2.2 adjusts them twice, i guess 0.4545 is the way to go?
When I adjust my linear image’s exposure to get rid of the burn, the whole thing goes darker. Does that suggest that I rendered with light sources set too bright in the 1st place? On the other hand if I render same settings but with Reinhard burn on 0.2 and untick “dont affect colours” the burn is instantly gone, the image nice and bright, but then I seem to loose saturation. I want to learn how to do propper LWF. What I’m trying to find out is where is my problem in my whole workflow. Is it my render settings, or is that part fine and there is a problem with my post work. I mean is it at all possible to get a correct LWF image, rich in colours and all the other wonderful things that a good image is suppose to have, straight out of render without having to do anything in post? Or is the render only 20% of the picture, and all the other bells and whistles gets done in post? Yes sure I could get good images without LWF, but the whole goal here is to learn to do it properly WITH LWF so that maybe one day some doors could open at larger studios that do use a LWF.
@DVP:
It certainly is possible to create a nice looking image without any postwork. The issues you mention suggest that the brightness of lightsources relative to each other is indeed off. All that the reinhard setting you mentioned does is exponentially adjusting the luminance. You can do that in post to using curves in PS for example to tone down the highlights while keeping darker areas at the same luminance level.
@pailhead: When you set the max Gamma/LUT settings to 2.2 with input and output set to 1.0 then you are rendering plain linear when saving via the VFB. You should tick “dont affect colors” and set the gamma to 2.2 tho, because that will force the sampler to do the importance sampling in sRGB space. If you dont you will unnecessarily need to oversample the image to get rid of noise.
Regards,
Thorsten
Right, i used to have a lot of trouble with anti aliasing when i started using lwf a few years ago. Adaptive subdivision was just too low in the shadow areas when using 1.0 gamma.
Still, it seems that i have to save an .exr image with .4545454545 to get a linear image in photoshop. 1.0 just gives a normal srgb image (the same thing i see with srgb preview in vfb)
Well i am not a PS user, but i assume that PS identifies the image as gamma 1.0 and autoconverts it to 2.2 ?