Photoshop save for web color shift
Baldy
Registered Users, Super Moderators Posts: 2,853 moderator
I think this is a simple question that should have a simple answer. We get the following question often and don't know what to say:
"I got my new Vista machine and all was well. Photoshop's working space is set to sRGB. Photos looked good in Photoshop and on the web. But seeking perfection, I hooked up a hardware calibration device, chose gamma of 2.2 and white point of 6500."
"Now, I see a big color shift when I choose Photoshop Save For Web. Photos on the web no longer look good or like my prints, only in Photoshop. What happened?"
I think I've read most of the color management books, but none that I've found answer the question. Can't seem to find it on the calibration vendor's sites. The color gurus we query say, "you have to view your photos in a color-managed application to be accurate." Unless I'm missing something, they haven't answered the question.
Can you?
"I got my new Vista machine and all was well. Photoshop's working space is set to sRGB. Photos looked good in Photoshop and on the web. But seeking perfection, I hooked up a hardware calibration device, chose gamma of 2.2 and white point of 6500."
"Now, I see a big color shift when I choose Photoshop Save For Web. Photos on the web no longer look good or like my prints, only in Photoshop. What happened?"
I think I've read most of the color management books, but none that I've found answer the question. Can't seem to find it on the calibration vendor's sites. The color gurus we query say, "you have to view your photos in a color-managed application to be accurate." Unless I'm missing something, they haven't answered the question.
Can you?
0
Comments
VIEW>PROOF SETUP>?????
I had huge problems w/ this before and worked through all the issuesI had. Hopefully I can help out!
Photoshop is previewing the numbers correctly. ANY ICC aware application is also previewing the numbers correctly (including an ICC aware web browser).
A non ICC aware application is NOT previewing the numbers correctly. Where's the confusing?
ICC aware applications USE the display profile to produce a preview. Non ICC aware applications don't.
What version of PHotoshop? Save for Web in CS and CS2 don't automatically embed sRGB in the doc's (CS3 does).
Author "Color Management for Photographers"
http://www.digitaldog.net/
I'll leave it to Andrew the Color Guru to answer why, but I do know that over the years most PS threads say to use the "Save As..." function rather than the Save for Web. Mainly to embed EXIF data but I believe there were color issues also. I just convert my color space in PS to sRGB and save as a jpeg file.
-Fleetwood Mac
That person is willing to buy your book and a calibrating device. If they calibrate and discover that pages on the internet are further from what they should look like than they were before, that's a very unexpected outcome.
It just happened to Andy on his Mac and Steve Cavigliano on his Vista machine, two people who are asked often to help our customers solve this.
If the answer is, "The monitor profile that was created is bad," I can at least understand that calibration does not break the internets. Then we can spend time trying to figure out if the room wasn't dark and light leaked in, or there was interference from a pre-installled profile, or something.
But if the answer is, "Internet browsers are not ICC aware so the Internets can look more busted after calibration than before," then I are confuzzzzzd.
That's just the nature of SFW though, since it's designed to optimize for web use.
There's also a cool little globe icon button in the bottom right corner of the SFW dialog box. (see screen grab) View the image side by side w/ the same image that you previously saved for web. You should see that there's no difference between images. If that's the case. SFW is doing it's job and your expectations of it are too high for this wonderful yet very focused app.
Unless your test come up different than mine. Your best answer may be to, "stay away from SFW" and just have a write up on what your findings are jsut like that great writeup on sRGB color spaces you guys did
There's nothing about 72ppi that would strip out any color. In SFW you can choose to keep all the colors (high quality JPEG or PNG) or strip out lots of colors (8-color GIF or low quality JPEG).
In Photoshop CS3 the number one cause of color shifts and "lightness" (contrast shift actually), has been solved. If you click the little round button to the right of the Preset menu, there's a new command, "Convert to sRGB" on that pop-up menu. As long as that's selected, images coming out of Save for Web viewed in browsers should be consistent with how they look in Photoshop. Although in some scenarios you may need to also use the Include ICC Profile option. But the key to everything is always converting to sRGB for the Web. Without that step, nothing else you do, Proof Colors or whatever, is going to help.
How can an image that's 400ppi get converted to 72ppi and still have the same colors? At that point you have to be averaging colors and sampling them together. If not, how does that work?
Oh, well, now that you put it that way, there will be some color shifting among local pixels. If you're talking about pixel averaging, then I agree that it's happening at that level. But I think that if it's making the entire image shift or change brightness overall, that's something different.
In Ted Padova and Don Mason's book, "Color Management for Digital Photographers for Dummies," they make an interesting point: unless you manually adjust your monitor to get it close before running a calibration, you probably won't get as good a result from calibration. I wasn't sure why that would be. Does the monitor profile only have limited ability to adjust the display color via the graphics card?
But the 30-inch Dells we are attempting to calibrate have no manual adjustments other than brightness.
And my confusion remains: why would the Internets look worse with a new custom profile than they did with the generic default one Dell provides? Andy's machine is a Mac, Steve's a Vista-based PC. Same monitors, same result: the new profiles made the Internets redder and more saturated. Andy did his calibration in New York with one puck, Steve in California with another.
I can't put a metric on what "pretty well" means. I can tell you that a set of RGB or for that matter CMYK numbers define a specific color appearance. In ICC aware applications on machines of users who calibrate and profile their displays, everyone SEEs the colors the same way. That the colors may not appear to some or all is immaterial, the numbers are previewing correctly, identically and consistently. The only way for this to work is for users to have ICC aware applications AND calibrate and profile their displays.
One could ask, "If you think the image you're now viewing on the web looks too dark, why not just increase the brightness of your display"? And yes, that would make the image look better. That no one else would see this might be an issue (it certainly is for people who hope to show their images to others).
There's better and then there's correct. Correct may not be better but at least the application isn't lying to you.
That the same numbers WILL appear the same with an ICC aware browser and the appearance correctly defines the numbers (that's all computers understand, numbers) what's the problem? Many here seems to want to defend stupid web browsers that lie to you about the color images they produce.
You solve it by using a web browser that treats your RGB images like your image processor.
Its possible for a profile to be "bad" but that's rare and not what's happening here. The display is in a fixed condition and defined by the profile but without the profile, the dumb web browser is even farther from the truth, its lying even more. What's broken is your web browser. I don't know why this is so difficult for many to understand.
The web browser is broken, get over it. Would you be happier if Photoshop was lying to you? Or Lightroom, Bibble, Capture 1, or just about every application on the Mac OS?
Its not like you have to shell out huge sums of money for Safari (its free).
Bottom line: IF you care about color appearance, then you don't want to use a browser that takes RGB values and previews them to you incorrectly.
Author "Color Management for Photographers"
http://www.digitaldog.net/
Based on that statement, the book probably is OK for dummies.
That's a really broad statement and since I don't have the book, I'll refrain from further saying, it depends.
On a CRT, you have physical adjustment over the behavior using RGB electronics. On an LCD, the ONLY control you have is over the intensity of the Fluorescent lights. There's nothing more you have to "adjust". Nor should you. Now if you're lucky enough to have a really high end LCD like my NEC 2690, it does adjustments in high bit (more than 8-bits) internally, NOT in the 8-bit graphic pipeline (doing so just adds banding to the display).
The calibration isn't all that important to ICC aware applications, the profile accurately defining the behavior IS. The profile does the adjustments in ICC aware applications such that multiple users all see the numbers the same way even though their displays absolutely do not behave the same way. And that's one reason why non ICC aware applications will look farther from the ICC aware applications, they don't understand the display profile and don't use it to adjust (compensate) so that identical RGB values produce the same color appearance on multiple users displays.
Say you're running Windows and I'm running a Mac and we both calibrate to a different Tone Response Curve (what most incorrectly refer to as Gamma). I'm at 1.8, you're at 2.2. ICC aware applications don't care. As long as your profile defines the display at 2.2 and mine at 1.8, the same set of RGB numbers preview the same on both our differing displays.
NO CCFL LCD does.
You're asking the wrong question. The question should be, IN AN ICC AWARE application, do they look the same.
Author "Color Management for Photographers"
http://www.digitaldog.net/
It seems like the only conclusions are:
1. Manufacturers should do their best to ship computers that display the Internet standard correctly: gamma of 2.2, white point of 6500, sRGB. Seems like Microsoft and the manufacturers of consumer electronics like televisions have this right.
2. Makers of calibration hardware and software should make sure your resulting calibration adheres to the same Internet standard. If, after a calibration, you need to have Safari and jpeg images with ICC profiles attached to see correct colors on your display, that's color mismanagement and not what most people are expecting. Am I wrong?
So you think all TV's display the data the same? Go into a TV store. All the TV's are getting the same numbers. Do they look the same? Nope.
We should cripple displays for the Web? I don't think so.
Displays are unstable devices. The same RGB numbers previewed today will not look the same in a year, or two years. That's why we calibrate them (and then profile them).
The internet is a huge wasteland, you have zillon's of users all working with different quality devices, with displays all in differing conditions. When the majority of web users feel about the consistency of images that those of use working in an ICC aware application like Photoshop do, we may get somewhere.
There IS no standard. And if people used the products above that you mention, it wouldn't matter. Its far, far easier for those who care about consistent color to simply use the right browser and of course, do what they've been doing to work in Photoshop and other ICC aware applications: Calibrate and profile their displays. Every month. More often with a newer unit.
No, your wrong. It works, it has since 1998 under Photoshop. Longer than that with other applications. You're making this far more complicated than it needs to be.
There are two groups at play here. Those who don't care what they see and those that do. Those that do use ICC aware applications and calibrate and profile their displays on a regular basis. Its worked well for a decade. Those who don't care (the vast majority of web users), well they don't care. Occasionally they return some product they purchased on the web because what they got and the colors they saw didn't sync up. Otherwise, they don't care. IF YOU CARE, you know what you have to do to fix the issue. Its simple and it works. We don't need (and I'd submit we don't want) displays trying to produce some standard, that hasn't worked from day one with the exception of very expensive reference devices (Barco, PressView, Artisan). If you can afford a $5K monitor, you certainly care about color consistency and you're not using Internet Explorer!
Now, what is going to happen in say 5 years (or less) when more and more users are working with wide gamut displays, those that approximate Adobe RGB (1998)? All the sRGB images on the web in non ICC aware browsers will look awful. All Adobe RGB (1998) images will look OK. That's happening more and more (the $5K wide gamut LED of two years ago is now $1800. An NEC 2690, "93%" of Adobe RGB is like $1200. Everything looks fine on Safari and like do-do on a non ICC aware browser. The sRGB color space is going to be history, a dinosaur very soon.
Author "Color Management for Photographers"
http://www.digitaldog.net/
I think it's suspicious that both machines were calibrated with Eye-One version 1 devices. Andy has the latest Eye-One coming today and is going to repeat the calibration. Maybe this is as simple as those devices are 4 years olde.
If you were to calibrate your TV and it no longer displayed your favorite shows as accurately as it did before calibration, I think you'd conclude the calibration was foul. I can't understand why the Internet should be different.
They all being the TV's in the electronics store? I'd take issue with that but I guess we'd need to agree on the store, the number of sets etc.
Yes, we had a long phone conversation about this yesterday. I don't suspect the i1 is a problem and there IS a diagnostic utility that could tell us if the instrument is an issue.
What do you mean by the word accurate? A set of RGB numbers in a known color space should produce a known color appearance. That doesn't mean anyone will "like" it. You're confusing consistency and preferences in color.
As for the Web, it IS correct on my browser. Its not on yours apparently. The images I see in Photoshop, corrected to produce the color appearance I desire and believe to be correct look exactly that way in Safari. That's why the internet is different for you, but not for me.
Are you actually telling me that images you view in Photoshop are not desirable but when you view them incorrectly on the web, you now like them and therefore, Photoshop is wrong and the net is right? Because if so, you're confused by the role of these two applications (Photoshop and a browser). The numbers you see in Photoshop are correct. If you don't like them, change them; that's what Photoshop is designed to do. If you want the web to match that, you need an ICC aware browser to match.
I don't know why you insist in going around in circles when this is very simple stuff. I don't have the time for it.
Is Baldy the only one here who isn't getting the concept? If so, I need to move on.
Author "Color Management for Photographers"
http://www.digitaldog.net/
Juergen
Medicine's mantra is, "First, Do No Harm." I don't see why the end result of calibration should be inaccurate colors in Firefox (and all but .01% of images be inaccurate in Safari).
Ahhh, this is where I was confused when I had this issue.
If you have an LCD display, you may NOT be calibrating your display to a particular color standard.
You may be just profiling your display (measuring it's current color performance and characteristics) and dropping that profile in a standard location for color aware apps to find it. Firefox may not display any differently before or after profiling your display.
Photoshop will find that profile and use it to modify it's color display according to the display's profile to get your display to show accurate colors. Firefox, since it isn't color managed, just sends raw data to the screen and you get whatever the monitor produces from the factory. For LCDs, where there really isn't much true calibration that can be done (other than brightness), running one of these color calibrators/profilers just doesn't do much for Firefox other than make sure you get the brightness into the right range (which is helpful on LCDs because they usually ship too bright).
FYI, CRTs were different in this regard and they could be "calibrated" somewhat so that their default color display (for non-color-managed apps) was affecting by the calibration/profiling operation.
Homepage • Popular
JFriend's javascript customizations • Secrets for getting fast answers on Dgrin
Always include a link to your site when posting a question
Yes, you've lied to Photoshop, you've circumvented its ability to properly show you the numbers correctly. Do you feel better now? You've gone out of your way to sabotage color management in Photoshop. This doesn't make it now correct (it isn't) but it matches the incorrect preview in IE or whatever stupid browser you're using. So now neither application is previewing the data correctly.
In Photoshop 5.0, there was a setting that turned off Display Using Monitor Compensation, effectively doing the same thing. People used it, forcing Photoshop 5 to act stupid like Photoshop 4 and earlier (and other non ICC aware browsers). Adobe was smart enough to remove this hurt me button in subsequent updates.
Look, you can set your soft proof in Photoshop (customize proof setup) to Monitor RGB. Now it will match (incorrectly) your dumb web browser. Feel better? Now its showing you the numbers incorrectly.
Again, with all due respect, you don't get it. You want to force Photoshop to incorrectly match a web browser that by default, doesn't preview the numbers correctly. What good does that do you? Nothing. Instead of wasting our time here making Photoshop preview the data incorrectly, or installing the wrong display profile, why not just get a damn browser that works correctly?
No, no no! Repeat after me: Safari is correct, Photoshop is correct. Firefox is not correct. Why do you keep insisting that the wrong preview is the one you want evey application to mimic? Why do you keep hitting your green head against the wall?
Author "Color Management for Photographers"
http://www.digitaldog.net/
Its ironic that people get the incorrect impression that they are calibrating their display to sRGB. Maybe, if you've got a circa 1993 CRT with P22 phosphors.
They don't understand that sRGB isn't much of a standard when you consider its NOT based on any real world let alone modern display. Its based on a theoretical color space that circa 1993 might be possible IF you followed proper calibration (rare at the time) and if the display was in a specific ambient light surround, at a specific luminance (one way too low for any LCD to hit).
Its easy to verify too if you have tools to view ICC profiles in 3D. Calibrate your display as best you can, view the ICC profile of it next to sRGB. If they are identical gamut maps, I'll buy you an expensive dinner anywhere you like (they will not be).
Then there's the fact that we calibrate displays on a regular basis because they do change behavior over time. Some a great deal!
If you go back to a circa 1993 expensive reference display, the Radius Pressview, a unit that could be calibrated with (at the time) very expensive hardware, you'd find that it did not produce sRGB. It produced when properly calibrated, ColorMatch RGB. You ever hear of this working space? It was around YEARS before Photoshop had color management and working spaces. When you properly calibrated a PressView to its target aim points, you did not get sRGB, you got ColorMatch RGB. Now the two are not hugely different in terms of color gamut but they ain't the same! And you had to drop thousand of dollars (circa 1990 dollars) to get a reference display that always produced the same color (ColorMatch RGB).
Some of you guys need to get past this idea that you can get your displays to produce sRGB. You can't. And you don't need to. All you need to do is fingerprint the behavior you get with a profile. And let that profile be used to properly produce a preview. What, you don't have an application that has a bloody clue how to use the profile? Too bad. You don't have sRGB and in this case, you have incorrect previews.
FireFox (at least the version that doesn't work with profiles) doesn't have a clue the profile exists or for that matter what a profile is. Just as if I write to you in German and you don't understand that language. Doesn't matter if what I'm saying is correct or not, you don't understand it. I can say your mother wears army boots or that you're the best looking person I've ever known, doesn't matter, you can't understand me. In this example, non ICC aware applications don't know a profile from a pickle. Get Safari or wait until FireFox understand what the profile is doing, we have a different situation.
Exactly!
Or whatever condition that display happens to be in (factory or otherwise).
Well yes, you had more control over a CRT but that's not an issue in this context. You can still muck around with LUTs on LCD's and make em look all kinds of different ways. CRTs did this using the electronics in the unit. So in this case, doesn't matter. What does matter, CRT or LCD is having a profile that defines the current behavior. Non ICC aware applications don't know what you've got in front of you with respect to what the display is doing. ICC aware applications do (if you've profiled the unit).
Author "Color Management for Photographers"
http://www.digitaldog.net/
Here's a useful link on what the calibrating/profiling software is doing:
http://www.pnwcmug.com/images/Monitor_Calib_Slideshow_22707.pdf.
It sounds like with LCDs, you can adjust the brightness (we all knew that) and by manipulating the LUT on the video card, you can set an appropriate white point. If you choose to keep the native monitor white point, then no adjustments in the video card are done. If you choose a particular white point, then some adjustments in the LUT may be applied. That, however, is not a fine grained color adjustment. The rest of the accuracy is achieved by profiling the result and then letting color-managed apps use that profile.
Here's a useful link that describes the different "calibration" capabilities of CRTs, regular LCDs and high-end LCDs: http://www.imagescience.com.au/ColourControl/colourProducts/calibrationAndProfiling.html.
Another good article on "profiling vs. calibration" and CRTs vs. LCDs: http://www.thinck.com/howto-1.html.
So, the big question here is "how much actual calibration happens on a given LCD?" No matter how much calibration is done, if an accurate profile is generated, then Photoshop (and other color-managed apps) will probably show accurate color results. But, if not much calibration is happening, then non-color-managed apps don't really benefit from the whole calibration/profiling step at all. And, in fact, what happens is that now you are much more likely to see how Firefox and Photoshop differ. Before you profiled, Photoshop and Firefox might have been the same (not accurate, but the same) so many would not have noticed.
My guess is that the profiling step leads to Photoshop becoming accurate, but now different than Firefox. The difference is now much more noticable than the previous inaccuracy was.
Are there any solutions here? The only ones I know of for your own system so far are:
- Use a browser that is color managed (Safari). Then, raw calibration doesn't matter since both browser and Photoshop are using the profile to correct the color display.
- Use a monitor that can be truly calibrated with the right software (for LCDs, I think this is only very high LCDs). That should give non-color-managed apps displaying sRGB data a chance at being decently close.
I'm looking around to see if there are any other practical answers. I don't yet understand what the limits are for LUT video card calibration on LCDs or what the various calibrator software packages even try to do on LCDs.Homepage • Popular
JFriend's javascript customizations • Secrets for getting fast answers on Dgrin
Always include a link to your site when posting a question
Or with far better and expensive units (NEC SpectraView, Eizo), you can do this in a higher bit depth within the panel itself (better).
Author "Color Management for Photographers"
http://www.digitaldog.net/
Andrew, I believe that the root of the misunderstanding here is that people expected a thing they bought to do color calibration to "fix" their monitor so the browser (even a non-color-managed browser, but displaying sRGB images) displays accurate colors. That's what I expected when I first profiled my display. You and I know it is more limited than that and only really solves the problem when using fully color-managed apps, but most people didn't expect that limitation or understand that limitation.
Here's a blurb from the i1Display2 web page:
"The award-winning i1Display 2 delivers unrivaled color controls including Workgroup Match, Ambient Check and Match, Push Button Calibration and Validation—all essential tools for professional photographers and designers to attain accurate color throughout the digital workflow, whether in their own studio or in a collaborative production environment."
While it doesn't say you will get accurate colors in Firefox, it's pretty easy to understand how people might think that would be a result of using the product.
Furthermore, I think we're seeing that the average digital photographer is much more likely to complain about their system if Photoshop shows different colors than Firefox (even though Photoshop is being accurate). There's nothing we can do about that, I guess - it' s just human nature until you get educated enough to understand what is really going on or switch to a color-managed browser.
I'm hoping that the real end-game here is a series of software improvements over time that make this issue eventually go away:
- Firefox becomes color-managed (we know they are working on this)
- IE becomes color-managed (we don't know if they are working on this)
- More color management support is built into windows
- Flash gets color management
- More of the web conserves ICC profiles in the images they serve
- Browsers support more efficient ways of tagging images with ICC profiles than having to embed a profile in every tiny image
- Safari switches to treat a profile-less image on the web as sRGB (or at least has a system preference for doing so).
In the meantime, I guess all we can do is look for work-arounds and help explain why it is this way.Homepage • Popular
JFriend's javascript customizations • Secrets for getting fast answers on Dgrin
Always include a link to your site when posting a question
Author "Color Management for Photographers"
http://www.digitaldog.net/
The link you post says the LED should use 'native' for the color temperature. What if that is not an option? What temp do folks use? I have seen 6000, 6200 and 6500, even occassionally 5000.
-Fleetwood Mac
Here's a recent discussion of that very topic: http://www.adobeforums.com/webx?128@@.3bc435af. Most of the folks in that thread (though it isn't unanimous) seem to say that D50 (~5000) seems to match the whites on photo paper the best, but I do also see the D65 (6500) recommendations. You can read yourself and see what you think. 5000 is typical "warm daylight". 5500-6000 is typical cool daylight. I'm amazed myself that there isn't one right answer to this question.
Homepage • Popular
JFriend's javascript customizations • Secrets for getting fast answers on Dgrin
Always include a link to your site when posting a question
I wonder if anyone saw this on the IEBlog re: IE8 and Acid2, plus the new IE8 Super Mode. What is interesting, since they are touting standards (specifically the W3C standards and the European Union's atttiude towards web standards), is that I can find no reference to ICC or Color profiles in those web standards? XML, HTML, CSS -yes. ICC - nada!
-Fleetwood Mac
After thinking about this some more, I wonder if the choice of color temperature for the monitor in the calibration/profiling process could be messing things up and actually making things worse for the monitor (with non-color-managed software) after calibration.
It appears that the one actual calibration step that can happen with regular LCD monitors is to manipulate the video card to set the color temperature. It stands to reason that if a wrong choice is made for the color temperature, then the default monitor display (w/o color management) could get worse rather than better. If that were true, then the inverse could be true also. It might be possible to actually tweak the color temperature setting until the non-color-managed display was pretty close to accurate (at least for one particular color range that you cared most about like caucassian skin).
Obviously, it wouldn't be as good a display as true color management because that requires more than just one linear variable (the monitor profile can be non-linear and vary by color channel). But, it should be possible to make it better. I might play with this on my monitor tonight. I also have a displeasing mismatch between Firefox and Photoshop for skin tone. If it's possible to tweak the monitor so this mismatch is smaller while still giving Photoshop an accurate monitor profile to work from, that would be nice.
Why we get to pick the color temperature or gamma at all still beffudles me. Why don't we just plug in the calibrator and tell it to just "do the right thing?"
Homepage • Popular
JFriend's javascript customizations • Secrets for getting fast answers on Dgrin
Always include a link to your site when posting a question
Because the "right thing" varies depending on what the heck you're doing.
If you want to keep a native setting, well use that if your calibration allows it (Native White Point and Gamma). Then you don't introduce any LUT, you simply profile the display as it is (which may or may not be "right" depending on what you're trying to soft proof). Same with luminance.
Author "Color Management for Photographers"
http://www.digitaldog.net/
This one has me confused.
Can you please explain why there isn't a single right answer for doing digital photography? When would the answer be 5000 and when would it be 6500 and when would it be something else? What do you choose and why?
Does the display of an image in Photoshop vary depending upon what color temperature setting in your calibration/profiling software you choose?
Homepage • Popular
JFriend's javascript customizations • Secrets for getting fast answers on Dgrin
Always include a link to your site when posting a question