Colors in 3ds Max

[David Arbogast]

I've really been struggling with the way 3ds Max displays color (as compared with the rest of Windows 7 and Photoshop). Whether it is the 3ds Max or Vray Frame Buffer or the "View Image" in the material editor, images display a distinct red cast in 3ds Max that isn't revealed in other apps on my PC; apps such as PS CS5 or even the built-in W7 image viewer.

 

My desire is to have images look the same across platforms on my PC...can anything be done to achieve that? I hope to learn what specific steps I can take (if any) to get all my apps (3ds Max, PS, etc.) to display colors the same and accurately.

 

[My system is a W7 64 bit PC with a Dell U3011 (wide gamut) monitor and I'm using Color Eyes Display Pro for calibration.]

 

BTW, why do 3d software providers, like Autodesk, fail to provide a color management system in their products...isn't it important to be able to make accurate assessments about image input and output within the 3d environment? Don't 3ds Max users, for instance, need to make accurate color decisions and assessments within the Max environment?

[Jeff Mottle]

Windows 7 and Photoshop are color managed, but as you know 3d apps (except Lightwave) are not color managed. If you had been working on an older CRT or LCD with a color gamut more closely approximating sRGB, these problems would be negligible. However because you are using a wide gamut display (which will be closer to AdobeRGB), your colors will shift a lot between the frame buffer and a color managed environment.

 

There is a trick I've developed which generally works quite well, although it's not something you would ever normally do outside of a 3d workflow. If you have done any research into color management you know that all of your devices need to be profiled and you need to assign those device profiles to your images when they are first opened in Photoshop. The exception is that you never assign a display profile to images. In this case however that is what you need to do. The logic is that your are making all of your decisions about color in the frame buffer based on the characteristics of the display. Therefore that display profile is the device profile you assign to the image.

 

Provided your profile is 2-3 weeks old at the most (VERY IMPORTANT), you will assign that profile to your frame buffer image when you import it into Photoshop. You then must convert it to a device independant color space before you start editing the image (as you always do when you import imagery where a device profile is assigned). The reason you do this is that device profiles are not linear in how they react to color adjustments. You could adjust red for example and at the same time be adding blue inadvertently. Depending upon how you have color management set up in Photoshop and what your working space is set to, you can either convert during the assign process or do it after the image is opened.

 

As all of the colors you are working with in the frame buffer are close to the AdobeRGB color space I would use AdobeRGB as your working space and ensure you convert into that space after your assign the image profile. If you will be sending the image for distribution to others outside of a color managed environment, I would recommend soft-proofing to sRGB, editing the image colors as required and then coverting to sRGB before saving a copy of the image out with the sRGB profile embedded. You could also use sRGB as your working space and save this step, but you have extra color bandwidth so prematurely cropping this color out before editing is a waste. Especially if you will be printing the images to an inkjet printer, and offset press or sending the imagery to another color manged environment or sent in a PDF (Acrobat is color managed)

 

As for why current apps are not color managed. Well there are a few reasons for this:

 

1) Most of these apps were developed for the gaming and VFX community where color edits are made in post, not in the frame buffer.

2) When the apps were first created most people used displays that had gamuts close to sRGB so there was negligible difference between frame buffer and Photoshop or other apps.

3) Making a 3d app color managed is not an easy task, and would require many apps to be completely rewritten to accommodate this.

4) It's never been a priority or a request by users.

 

This having been said, Autodesk has been making steps towards accurate color management within the app, as evidenced by the recent ColorMunki plugin for 3ds max. I think it will eventually be color managed, but it's hard to say when. XBR as many know is the complete rewrite of the core engine, so it's more likely to happen now than it ever has been. I've been lobbying for it for a number of years and they do know it's becoming more and more of an issue, especially as more and more people start using wide gamut displays. Years ago I predicted that as wide gamut displays became more common place this would act as the catalyst for color managed 3d apps. It certainly appears to be raising awareness of this issue. We will have to wait and see..

 

Anyway, hopefully the above helps. Let me know if you have any questions.

 

Cheers,

Jeff

Edited March 7, 2011 by Jeff Mottle

[David Arbogast]

Jeff,

Thanks so much for the detailed great response...it's exactly what I was hoping for and I'll give your suggestions a try.

 

Similar but different question: what color setting do you set your wide gamut monitor to (in the monitor's menu color settings). Adobe 1998? I'm away from my desk at the moment, but I'm wondering if having the monitor's menu set to Adobe 1998 isn't causing my colors to look wrong in Max. I am puzzeled as to what role the monitor's own menu color settings have in a color managed system.

[Jeff Mottle]

 

Similar but different question: what color setting do you set your wide gamut monitor to (in the monitor's menu color settings). Adobe 1998? I'm away from my desk at the moment, but I'm wondering if having the monitor's menu set to Adobe 1998 isn't causing my colors to look wrong in Max. I am puzzeled as to what role the monitor's own menu color settings have in a color managed system.

 

Most displays do not allow you to set on display color gamuts, but all that is doing is internally clipping the gamut of the display before it reaches the graphics card. My display does not do that. A couple of weeks ago I actually did an experiment and profiled a wide gamut display (down at Neoscape) normally and then with it's mode set to sRGB. I have special software to view 3d and 2D representations of profiles. The wide gamut display on it's own as expected produced a gamut close to AdobeRGB. When I set the display mode to sRGB and re-profiled the profile was now the same size as sRGB. It did what I expected, but I was curious to see the raw data for myself.

 

You could potentially try setting the display to AdobeRGB then calibrate and profile the display. You still have to do this. Then (in theory), your display should be showing EXACTLY AdobeRGB, so you would not need to assign the display profile, but rather just assign AdobeRGB right away and you are already in your working space. The less color space conversions you do to an image the better.

 

If you set it to AdobeRGB and do the calibration and profiling, send me the ICC profile and I'll import it into my profile visualization software. Then we can see how close the in-display gamut settings come to the actual AdobeRGB profile. Before you do, send me the profile of the display profiled in it's normal mode too. That way I can tell tell/show you what colors are being clipped (if any) in that mode from the native display gamut.

 

Also just to clarify, the reason I suggested applying the display profile first in my first post is that while wide gamut display closely approximate AdobeRGB, they are rarely exactly the same (barring our experiment above). So there could still be color shift between color managed and non-color managed devices.

[David Arbogast]

OK, I think I'm getting to a place of satisfying results. Here's what I've experienced thus far:

 

1. My original post was actually more driven by my frustration with how texture images appeared in 3ds Max vs. Photoshop. I had been re-coloring a beech wood texture in photoshop to have a sandy-blonde appearance, but in the max material editor ("view image") the sandy-blonde texture became reddish and fleshy looking. Very frustrating!

 

2. Tonight in Photoshop I tried converting and saving the wood texture to Adobe1998 color profile. The result is that the wood image looked exactly the same in Photoshop and in Max, which is great. So, any image that hasn't been assigned a color profile, or has been assigned to a different profile (sRGB or ProPhoto) and saved in Photoshop looks radically different viewed in 3ds Max vs. in Photoshop. So, my experiments have shown that if my texture maps are assigned/converted to the Adobe1998 profile, then I can obtain (to my eyes) a pretty perfect color match between Max and Photoshop.

 

3. Setting the Photoshop working color space to Adobe 1998 has also produced the pleasing result that the Vray Frame Buffer image looks like a close-enough approximation of the image after importing to Photoshop. I didn't experience any visual change by first assigning my current monitor profile to the image and then converting to working space (Adobe1998) versus immediately converting to working space.

 

So, things are seemingly looking good as of now, but I have one question/potential concern (hence "seemingly"): In order to replicate the above experience (#2) all new or existing texture maps need to be assigned/converted to an Adobe1998 color profile. That's easy to do in Lightroom (which I have, so no troubles there), but I just want to ask if the idea of converting/assigning my whole texture library to Adobe1998 raises any cautionary red flags.

 

Funny, I was entirely oblivious to the world of color space until getting this new wide-gamut monitor and started experiencing color-shifts that I hadn't experienced before. It has given me some bad brain-aches, but if a Photoshop-to-3ds Max-to-Photoshop workflow can be nailed down it will be well worth it! For one thing, it has caused me to invest in calibration software and equipment (thanks to your paragraph on High-Gamut monitors where you indicate that color management is a necessity with such monitors).

Edited March 8, 2011 by David Arbogast

[Jeff Mottle]

2. Tonight in Photoshop I tried converting and saving the wood texture to Adobe1998 color profile. The result is that the wood image looked exactly the same in Photoshop and in Max, which is great. So, any image that hasn't been assigned a color profile, or has been assigned to a different profile (sRGB or ProPhoto) and saved in Photoshop looks radically different in 3ds Max. So, my experiments have shown that if my texture maps are assigned/converted to the Adobe1998 profile, then I can get (to my eyes) a pretty perfect color match between Max and Photoshop.

 

Assigning a profile simply assigns meaning to the raw underlying RGB numbers in a color managed environment. By assigning a profile you are simply telling Photoshop's color management engine how to interpret the RGB values. When you convert to another color space you are actually changing the RGB values so that the color appearance is preserved when converted the destination space. Although what you are doing works, I would highly caution against converting your entire texture library. You will in some cases cause artifacts by converting imagery that almost certainly was created in a much smaller sRGB color space and making it stretch out in AdobeRGB. You could get banding on gradients and some unpredictable color shifts. If you are going to work this way, KEEP THE ORIGINALS as converting to another colorspace is not reversible. Also anyone else who tries to work on your file on a different display will get different results and you will run into issues as newer displays increase their gamut.

 

3. Setting the Photoshop working color space to Adobe 1998 has also produced the pleasing result that the Vray Frame Buffer image looks like a close-enough approximation of the image after importing to Photoshop. I didn't experience any visual change by first assigning my current monitor profile to the image and then converting to working space (Adobe1998) versus immediately converting to working space.

 

Yes, that is because a wide gamut display is close to the AdobeRGB color space. You will likely still see some shift, but how much depends on the particular characteristics of your display.

 

So, I think I'm pretty happy with how things are looking, but I have one question/potential concern: To replicate the above experience (#2) means that I will need to assign/convert any new or existing texture maps to Adobe1998 color space. That's easy to do in Lightroom (which I have, so no troubles there), but I just want to ask if the idea of converting/assigning my texture library to Adobe1998 raises any cautionary red flags.

 

See my note above.

[SgWRX]

this is something i was just looking at again. i tried to make a gretag color checker inside max. i use mentalray by the way. i converted teh sRGB colors from gretag to a&d diffuse color by dividing by 255 (i think this is correct from a post here a while back) but now i'm second guessing because of an article about daylight analysis with mental ray. in it they show the corresponding rgb color of patch neutral 5 .70D as .1912. when i do the rgb conversion i get .478. my conversion of black, gives me .20 which is closer. so maybe there is a curve i'm not taking into account.

 

also, i realized that even if the rendered frame window is showing a 16bit image and i save it as a 16bit tif, we're assign a profile of sRGB so what you're saying is those colors outside sRGB in the render are getting lost.

 

thanks,

[Jeff Mottle]

this is something i was just looking at again. i tried to make a gretag color checker inside max. i use mentalray by the way. i converted teh sRGB colors from gretag to a&d diffuse color by dividing by 255 (i think this is correct from a post here a while back) but now i'm second guessing because of an article about daylight analysis with mental ray. in it they show the corresponding rgb color of patch neutral 5 .70D as .1912. when i do the rgb conversion i get .478. my conversion of black, gives me .20 which is closer. so maybe there is a curve i'm not taking into account.

 

also, i realized that even if the rendered frame window is showing a 16bit image and i save it as a 16bit tif, we're assign a profile of sRGB so what you're saying is those colors outside sRGB in the render are getting lost.

 

thanks,

 

I thought about doing this myself, but it's not going to work. I've discussed this with Pierre-Felix (the guy at Autodesk who worked on getting 3ds Max certified for accurate lighting analysis) and it's not going to work as expected.

[SgWRX]

hm. i ended up finding a formula and updated my colors and it looks OK? but maybe not as exact as my 20% reflective black is still off compared to Pierre-Felix's paper. i found this formula: (161/255)^2.2 where 161 is any of the R G B values. I was pointed to a pdf file where someone coverted the CIE LAB colors to sRGB, aRGB and I think Prophoto rgb values.

 

their paper on daylight sim is great! i wish it were a book on the subject rather

[Jeff Mottle]

hm. i ended up finding a formula and updated my colors and it looks OK? but maybe not as exact as my 20% reflective black is still off compared to Pierre-Felix's paper. i found this formula: (161/255)^2.2 where 161 is any of the R G B values. I was pointed to a pdf file where someone coverted the CIE LAB colors to sRGB, aRGB and I think Prophoto rgb values.

 

their paper on daylight sim is great! i wish it were a book on the subject rather

 

Would be curious to see a step by step on this to review.

[SgWRX]

i downloaded this: http://www.babelcolor.com/download/RGB%20Coordinates%20of%20the%20Macbeth%20ColorChecker.pdf

 

pdf and reviewed (not too thoroughly though ) and ended up using the sRGB values from GMB (i believe that's the numbers gretag give).

 

i created a plane with enough segments to correspond to the colors, coverted that to an editable poly and applied a&d materials with a reflectance of 0 to each of the squares i ended up with. the squares are 2x2".

 

in order to get the color values for each of the 24 swatches, i first tried taking the r g or b values and dividing by 255. keeping in mind the apparent effect of indirect illumination and mrsun/sky i still thought they were too washed out. i turned off gamma correction in max (2011 design) and in the material editor, the colors looked more saturated and darker, or less pale.

 

i did various things such as saving an 8bit jpg and 16bit tif files with and without the gamma correction and using a photometric light then again with mrsun/sky and indirect illumination.

 

after more digging, i found a comment on mymentalray and it talked about a reflectance percentage (probably from PF, and i thought it looked similar but different to other formulas i've seen, but for some reason i kept getting #s that were >0. those were all too bright. so i tried this formula (mentioned above) and it seemed to work.

 

there's got to be some differences though. there's definitely a lot of places error can creep in. for example my R value for the cyan patch ended up being 0.000 because it wouldn't accept 0.0001 or something like that, not enough digits which i think can be set in preferences? anyway. what i want to do is go back and render under at photometric light only then bring into photoshop, assign the sRGB color space and see if the values for RGB are close to the values in the PDF file.

 

i believe they'll be off, but at least regards the reflectance values in Pierre-Felix's paper (i know it's co-authored but can't recall the 2nd name sorry), it's fairly close in the gray shades. i think the .1912 turned into .192 or maybe even .193. so right there is some error.

 

i'm going to do some more research to see if i can find reflectance values on the gretag, i'd like to know what formula PF used.

 

i do wonder what the "f-stop" range of a camera in max with mrphoto exp. control would be, now i might be able to figure it out. the next thing i want to do is a white/gray/black card would be. i believe the gretag color checker and the white/gray/black use the same white point, middle (18%) gray and black point.

 

i'm looking for this under photometric lights and then also under daylight conditions. the thought is, the color shift from the brightness and color of the sun/sky should be similar to real life. so you could color correct from there.

 

anyway, i'll post the scene tomorrow when i'm back at work.

 

thanks,

[SgWRX]

Here's my scene file. It's max 2011 design.

[SgWRX]

of course light is an issue as well as exposure. seems under my scene foliage was brighter than it was when i typed the srgb numbers in photoshop. purple was darker than when i typed numbers in photoshop. hm. perhaps thats the point though, where one would use this in real life with a camera, you'd run a script depending on your raw converter and profile the camera to a certain degree. in theory though you'd be able to reproduce the colors under the same lighting conditions.

 

i could check again to see if i can render without lighting effects and see if i can get the same colors.