Working in Non-Linear Space

[3ddesign]

Just as an FYI, currently there's a linear workflow discussion going on on the Fusion mailing list. The AIM-DTP website is evidentally full of misinformation and the way they calibrate the monitor isn't entirely correct.

I think the only trusted and repeatable way to set up is hardware calibration.

Yeah, I figured some of the AIM stuff was off a bit. The tests I made just didn't make sense. However, the guys in our graphics department tell me that the Adobe Gamma Calibration I did was quite accurate. At least I'm satisfied that my scans and renderings are printing a lot more accurately and match my monitor much better than before. I think that over time we will all learn a lot more about linear workflow, at which time I'll be the first to convert if it works.

Craig

[dapeter]

Hey - don't look at me I'm learning this stuff too... I saw all the stuff on the Fusion list, and then went and re-read all the stuff here. Made a closed room full of good ol' teapots scattered around, with a single light on top. Hit render... man what a difference! I was wondering what's been holding me back in the look I've been trying to achieve... I thought it was color mapping but it looks like it's the linear workflow thing.

Of course Viz 2007 has the minor glitch that you can't display material editor samples with gamma correction enabled... grrr.

[3ddesign]

When I alter the gamma to 2.2, everything (existing models and maps) looks washed out, so presumably you have to start a new approch to compensate by altering your lighting and maps.

Yes. I had to scrap a lot my my old texture maps due to lousy color rendition. Its amazing how messed up my materials were. I also reduced my lighting considerably. I was always increasing the multiplier to 15 or 20. Now its down to 1 to 2 and my shadows seem to be much better.

As a side note, the ImageCEL texture library I use renders much better straight off the CD than before, where I had to take them into PhotoShop and tweak the dickens out of them.

Craig

[3ddesign]

Hey - don't look at me I'm learning this stuff too... I saw all the stuff on the Fusion list, and then went and re-read all the stuff here. Made a closed room full of good ol' teapots scattered around, with a single light on top. Hit render... man what a difference! I was wondering what's been holding me back in the look I've been trying to achieve... I thought it was color mapping but it looks like it's the linear workflow thing.

Of course Viz 2007 has the minor glitch that you can't display material editor samples with gamma correction enabled... grrr.

What settings did you use for the teapot room? Can you post them as well as a render?

[flipside]

Nowhere in the LWF topic from throb or Gijs is suggested to calibrate your monitor to 1.0. They all say 2.2, that is why you will also set your max gamma to 2.2. If you would set your monitor to 1.0 then you should leave the max gamma also to 1.0.

What is referred to as linear workflow has gotten a broader meaning over time. Linear workflow means you keep your image data in linear space, but to view it correctly in gamma 2.2 space you must setup photoshop or whatever program you use. This should have advantages in manipulating images as suggested at aim dtp. If you don't care about these minor photoshop issues (like me), you don't have to work in linear space but you can still benefit from the gamma 2.2 thing. If you work directly in gamma 2.2 this means the gamma 2.2 curve will be burnt into the image (which is not the case in the real LWF). When I talk about LWF I actually mean that I use gamma 2.2 correction to my images.

So if you're only interested in producing nicer lighting without the usual tweaks to brighten up interiors for example, you don't really have to worry about all the calibration and profile stuff. All this is merely a side note if you wan't to produce extremely correct colors.

So in order to benefit from the gamma 2.2, all you have to do is what juju said in his sum up: (step 1 isn't even really necessary)

1) You should calibrate your monitor to 2.2 Gamma Workspace. This is independent of working in Max but relates to improving quality overall. Information can be found at the AIM-DTP website and is very worth taking the time to read.

2) Now for the workflow. Set 'Max Gamma' (under preferences) to 2.2. Set 'Bitmap Input' in same window to 2.2. Leave output to 1.0

-Bitmap input means that when you import bitmaps into your diffuse swatches, they will already be corrected to the new working space so that you don't need to use ColorCorrect plugin on these particular materials. This is your choice - Flipside likes to not set the bitmap input to 2.2 cause he likes to manually use Colorcorrect on all his materials. I personally think it's more work. *Just note that setting the bitmap input to 2.2 adversly effects HDRs/EXRs (if you plan on using them.)

3) Convert diffuse material slots to new workspace. It only takes a second or two for each mat. In each diffuse that you use, right-click your Diffuse vray color swatch and hit 'copy.' In the blank square next to it, click and choose ColorCorrect (a free plugin which you should install.) Paste the color you copied (right-click paste) into the the red ColorCorrect swatch. Set Gamma in this same rollout to 2.2.) Done.

4) Set 'Color Mapping' in Vray Rollout/Render dialogue to 'Gamma correction' Values should be '1 (dark)/.45454 (bright).'

5) Render your scene to Vray Frame Buffer (VFB), not the Max Frame Buffer. The VFB now displays the final image without having to change/add anything. If it is too dark, turn up your light multipliers and rerender.



This is actually what 3ddesign is doing too imo. He only does the whole calibration stuff with scanner and printer etc... Like said in point 1 of juju's mini tutorial, "This is independent of working in Max but relates to improving quality overall".


Old scenes should be fully converted to the new gamma 2.2 (color correct, lighting etc...). New scenes have no problem, you should only make sure that you convert your textures like explained (color correct plugin or bitmap loader). A hdri is usually 1.0 already, so doesn't need a conversion. If you set your bitmap options to convert by default, you should make sure you don't apply it to the hdri!

[flipside]

I also reduced my lighting considerably. I was always increasing the multiplier to 15 or 20. Now its down to 1 to 2 and my shadows seem to be much better.

Same here, and using lower multipliers means faster render times. You don't need tricks to brighten up the dark regions, which also means faster render times. You will also don't have the excessive color bleed you always have in the normal workflow. All simply by using gamma 2.2 correction, really amazing!

Here's an old project I just converted to gamma 2.2 rendering. From the moment you are happy with lighting balance, you can simply turn up the dark multiplier in the gamma correction color mapping, that is the same as increasing all your lights strengths. This can look much better of course, but this was a quick conversion. Note there is no green wall because of the displaced grass. In my old workflow, I had to put a vraymtlwrapper on the grass material to control the send GI so the color bleed was a bit less. Now I don't need any of those tricks. In the old one, I also used GI multipliers higher than 1.0, which now isn't necessary either (=a lot more logical!).

I don't think this image has no contrast or washed out colors...

[TRGraphics]

Changing the bitmap input to 2.2 made the difference here when using 2.2 gamma. Thanks!

Nothing like coming late to the game I see a big difference already in my print output! Very , very close match now.

I ordered a Spyder to calibrate my monitor. This should make it easier to get things perfect for the printer.

[dapeter]

I might be able to post the images at home. Basically it looks like the difference between the 'greebled' images on the tutorial:

http://throb.net/site_main/LinearWorkflow.html

Viz has a different dialog than Max - you can enable the correction, and enable the bitmap in/out, but there's no checkbox for the material editor. All I did was gamma correct to 2.2. I also left the bitmap box alone (so 1.0 in, 1.0 out), so when I load a bitmap I override the gamma to 2.2 in the loader dialog box.

I also use the color correct plugin on any textures that are derived from pure RGB data in models that are already done and I'm reworking them.

The only disappointing thing is how long I've been working the 'other' way without even knowing about this... I've done a ton of research on Max/Viz/Vray/MR/etc... but not too much on color management (even though I'm aware how critical it is).

[dibbers]

If you leave gamma output to 1, won't that save and look differently in PS to how it views in Max?

[3ddesign]

4) Set 'Color Mapping' in Vray Rollout/Render dialogue to 'Gamma correction' Values should be '1 (dark)/.45454 (bright).'

I've tested using the Gamma Correction color mapping (using your settings) and the Linear Multiply at 1,1. The following images show how close the two are. However, if you look at the border between, say, the magenta cube front right and the cyan cube back right, there is a dark AA irregularity with Gamma Correction but looks normal with Linear Multiply. I'm wondering if this is because the texture already had 2.2 gamma burned in, then if you render with gamma correction again, it is getting a double correction? Does this make sense?


gamma correction color mapping:


linear multiply color mapping:

[Gijs]

@ 3ddesign,

All the things you describe is exactly what is called working in linear space. I don't see a single difference. In the tutorial on my website I described two ways to work linear:

use gamma correction color mapping
use linear color mapping and assign a linear profile

IMO the last method is exacly what you are using. please indicate at what point do you think it differs, because I couldn't find it

[flipside]

Not sure what you are doing here. Is this a scene or is it a BG image you're rendering?

There are 3 ways of getting gamma correction:

1. use gamma color mapping as said
2. don't use it and apply throbs curve to the vray frame buffer so your image looks correct in the VrayFB
3. don't use vray frame buffer. The max frame buffer is gamma 2.2 corrected because of the display settings so your image will look correct(vray frame buffer isn't affected by the display gamma settings!)

The difference between 1 vs 2 and 3 is that in 1 the gamma 2.2 is burned in. In 2 and 3 it is not and if you open that image in photoshop, you need to apply a correct profile to it otherwise it will look very dark.

This also means that in 2 and 3, the image isn't gamma corrected at rendertime (it is done in the viewing system: throb curve, max display gamma, PS color profile). So it is in fact a pretty dark image. Vray will sample dark areas not so good, resulting in bad AA for example.

This image was created with option3:



This one is the same, but rendered with the vray vfb and gamma color mapping (1/0.455). (option 1)


Notice the difference in AA quality!

The first one rendered faster though, because vray didn't sample that much. So if you use option 2 or 3, you need to increase the QMC sampler settings to get clean enough AA.

[jow]

I also reduced my lighting considerably. I was always increasing the multiplier to 15 or 20. Now its down to 1 to 2 and my shadows seem to be much better.

Same here, and using lower multipliers means faster render times. You don't need tricks to brighten up the dark regions, which also means faster render times. You will also don't have the excessive color bleed you always have in the normal workflow. All simply by using gamma 2.2 correction, really amazing!

Here's an old project I just converted to gamma 2.2 rendering. From the moment you are happy with lighting balance, you can simply turn up the dark multiplier in the gamma correction color mapping, that is the same as increasing all your lights strengths. This can look much better of course, but this was a quick conversion. Note there is no green wall because of the displaced grass. In my old workflow, I had to put a vraymtlwrapper on the grass material to control the send GI so the color bleed was a bit less. Now I don't need any of those tricks. In the old one, I also used GI multipliers higher than 1.0, which now isn't necessary either (=a lot more logical!).

I don't think this image has no contrast or washed out colors...

Looks good to me
I think my problem is i always try LWF with exterior scenes that are already setup with standard lighting settings but with out lowering the sun/sky multi.....

Sorry to hijack but i had a little play with a scene earlier and converted it to LWF and came up with this....


the same tree,2 angles. Vray mats on the tree.
If i put a HDRI in the enviro slot i get Black spots Every Where
Any idea why?

[3ddesign]

@ 3ddesign,

All the things you describe is exactly what is called working in linear space. I don't see a single difference. In the tutorial on my website I described two ways to work linear:

use gamma correction color mapping
use linear color mapping and assign a linear profile

IMO the last method is exacly what you are using. please indicate at what point do you think it differs, because I couldn't find it

It is probably a semantic thing, a confusion of words rather than of technique. I was under the impression that Linear meant working with images that had a linear response vs. a curved or gamma response. If this isn't the meaning of working in linear space, I guess you won't find much difference.

I began this thread after becoming totally confused with all the discussion of LWF. Most of it didn't make sense. I wanted to clarify for myself and others, who might also be confused, what this was all about.

For example, the AA effect in the post above at least raises questions about the effectiveness of using Gamma Correction in the workflow I outlined. From the responses on the LWF thread, I think too many people may have just accepted the information presented by the Vray Gurus (for whom I have nothing but total admiration, so don't think I'm speaking negatively) without taking time to test the results. So I created a simple test and posted it as a way of moving the dialog further. I really hope I didn't sound like I was being critical. I was just trying to understand.

Craig

[dibbers]

The difference between 1 vs 2 and 3 is that in 1 the gamma 2.2 is burned in. In 2 and 3 it is not and if you open that image in photoshop, you need to apply a correct profile to it otherwise it will look very dark.

So why don't you output to 2.2 gamma as well then?