Web browsers, profiles and the “Color Appearance” tool

[alicem]

TL; DR

• My new HP laptop’s Xubuntu has a weird “Automatic - Built-in display” color profile enabled by default in the system settings, for the internal monitor.
• For some obscure reason, when this profile is enabled, it seems to be applied in Firefox to every image that has an embedded profile (but not to profile-less images), even when displayed on an external monitor, and this is the only change I notice on my system when toggling this profile on and off.
• Keeping this profile off allows me to see the same thing in Firefox and other apps.
• The “Color appearance” tool is probably not to blame at all, but I should perhaps refrain from doing overly fancy things with it, especially the “Viewing conditions” part.

Original post

Hi. Sorry, this will be a somewhat lengthy post, but I’m being driven insane by color stuff again. I did my best to highlight the main questions with quote markup and gave them numbers. Well, they’re more like groups of questions, but still…

For some reason, after putting more than twenty series of pictures online over the course of multiple years, it’s only now that I realize that they look different in Firefox than they do in RawTherapee, in my image viewer (Ristretto), in Chromium, and in a GIMP import. Basically, all those apps show me impossible-to-distinguish results, except Firefox, which yields warmer and slightly more contrasted results, be it for JPGs or PNGs.

I then looked around and stumbled on Web browsers color management (solved). To be perfectly honest, beyond the few things that I learned to deal with camera profiles in [Questions] How to choose a DCP profile (if at all) (after which I just basically stuck to Adobe’s “Neutral” profile), anytime I’m dealing with such topics, I feel like a toddler trying to lean quantum mechanics. Despite how perfectionist I can be on many RawTherapee settings, I never got the motivation to properly delve in that color profiles stuff. I just noticed that my pictures looked satisfactory to me on my screen, on my parents’ screen, on my mobile phone, and that people were generally happy. So the last thing I needed was yet another dive in what has been poetically described as…

@Claes

… a combination of an abyss, a quagmire, and a jungle, with a can of worms admixed.

The aforementioned thread about web browsers, even if mostly cryptic to me, allowed me to notice that Firefox’s gfx.color_management.native_srgb setting (from about:config) seems to be the main culprit: it is false by default, and setting it to true makes Firefox “agree” with the other apps.

Hidden details about caveats when fiddling with this setting

• I need to close the browser tab and re-open the picture ; a refresh, even a “hard” one, does not seem to do the trick.
• Probably due to weird Firefox optimizations, it seems that trying to open the same picture simultaneously in two tabs such that one has native_srgb = true and the other “native_srgb = false” fails: the config change is ignored for that picture as long as the picture is still open in one tab, so it looks the same in both tabs no matter what I do. (This is not a real issue but it made my analysis harder until I understood this).

Expandable details about my system and my very basic, very default-ish settings

RT 5.12:

Xubuntu 24.04:

Firefox 142.0 (snap):

Question 1

What’s the deal with that native_srgb setting? I read on the web that “It hands off color management to the system” (when true, I suppose). I don’t get why it’s not true by default, then. Perhaps that changed recently – that would explain why I did not notice the differences until today. What’s the point of letting the browser do weird stuff with our colors? Unless letting Firefox handle this (false) allows it to parse the embedded profile, but that would mean that Firefox’s weird warm display is the most accurate one and that _all the other apps, including RT, have been misleading me?

I initially thought that it was only happening on my newer pictures, processed with RawTherapee 5.12, but it seems that it’s actually just that it’s more visible when there are orange-red-ish colors (and perhaps blue skies too?). This matches the following observation:

@XavAL

In the image I chose as an example there is a good range of yellows and oranges, and maybe because of that I saw the differences so clearly.

The rust on top of that RT5.11 pic becomes distinctly more vibrant on Firefox (at least on my machine…), for example. Perhaps the phenomenon became worse with 5.12 but that would be weird.

Beside the RawTherapee upgrade, though, there is a thing that changed in my workflow: I started using the “Color Appearance” tool from the “Selective Editing” tool – generally with a “global” spot, but also sometimes locally, and occasionally both at the same time. To make matters worse, after noticing that applying the “Color Appearance” tool seemed to make my colors colder, I took the habit of “fixing” this by slightly bumping up the “Chromatic adaptation/Cat16” value (no further than 2 or 3, but still) in the “Viewing Conditions” part of “Color Appearance”. One thing leading to the next, I also often fiddled with the other sliders of “Viewing Conditions”.

Question 2

Could “Color Appearance” (in a global spot and / or locally) be making the “looks different in Firefox” issue even worse? I’m getting this impression, even with the default “Viewing Conditions” settings. But it seems even worse than worse when I fiddle with the “Viewing Conditions”. Am I imagining things? Should I go over my whole recent picture series (on which I spent dozens of hours already…) and revert some of those settings before putting anything online?

Expandable comparisons

No “Color appearance”: Firefox looks a bit more vibrant and contrasted, but it’s not really a bother:

Intentionally ridiculously high value for viewing conditions: The pictures look distinctly different to me:

Honestly, I was not sure what the “Viewing Conditions” settings are intended for – I assumed it was meant to adapt to specific viewing conditions, like when physically exposing pictures in a place with suboptimal or unusual lighting conditions (RawPedia seems to confirm that, if I understand it correctly), but I did not think it would have any adverse effect if I used it like some kind of hack, as long as my white balance made sense. Kinda like color toning or the RGB curves, so to speak. Perhaps I was utterly wrong and made a big mistake there – on dozens of pictures.

Question 3

Should I stop fiddling with the “Viewing Conditions” settings if I’m not trying to adapt to… well… real viewing conditions? But why does “Color Appearance” keeps making my pics slightly colder? Am I supposed to systematically counterbalance that with a “Warm / Cool” tool in my global spot instead?

Question 4

I surmise that there’s generally no such thing, in the absolute sense, as being “wrong” or “right” regarding colors, but still:

Who should I trust when optimizing my raw edits? For example: If I go to a serious, professional print shop and hand them a PNG or TIFF or even JPG exported via RawTherapee 5.12, and if their software takes the embedded RTv4_sRGB profile into account, probability-wise, what will the result be closer to between:

• Team Firefox with “native_srgb = false”, and…
• Team Image Viewer, RawTherapee GUI preview, Chromium, and Firefox with “native_srgb = true”?

Question 5

When people say stuff like…

@Jossie

In Germany almost all printing services do not process embedded ICC-profiles. They just assume, everything is in sRGB. So if you render the output to sRGB colour space you are on the safe side.

• … Does RawTherapee’s RTv4_sRGB count as “an sRGB that puts me on the safe side”?
• Does the meaning of “rendering”, here, fit what is done when exporting from RT with a profile that gets embedded into the exported image, or does it mean converting the image into a profile-less one by applying the sRGB profile once and for all to its pixels?

Finally, perhaps the most naive and stupid question of all – please don’t laugh or scold me:

Question 6

Why embed profiles into exports at all, if they seem to cause issues like that? Why wouldn’t the file just toss RGB values for each pixel and call it a day? Why embed something that makes the picture’s look depend on the browser, its settings, and even possibly the OS?

I did a very stupid test: I imported my file in GIMP, let it convert the colors from the embedded profile to its own profile (I don’t see much of a difference between what “Keep” and “Convert” yield), and exported the image as a new file while un-checking the “Color profile” box, so the new file was profile-less. And guess what? This profile-less file has a consistent appearance across apps…

Expandable comparison video

Not sure the .webm will do it justice, but here’s me switching between two Firefox tabs. The version with the red-ish sky and not-quite-black shadows in the tree is the one with the embedded profile (the “oddly warm” version), while the version with the more cyan-like sky and contrasted tree is the GIMP-produced profile-less version, which looks like what my other apps display:

Here’s a low-quality version of the picture (but perhaps the server will perform weird conversions on it that will remove the profile? Nope: If I download it, it still has the profile):

Seriously, I’m now considering trying to script this process in order to upload profile-less JPGs on my website. Or to ask RT to export with no profile, but I’m not sure at all that it’d be equivalent and harmless. Or just return to not caring about this, especially considering that my family and friends will look at my pictures on all kinds of devices, monitors, operating systems, and lighting conditions – but for this I would still like to know whether the most accurate representation of my picture is Firefox’s weird warm version or the other one. And whether the “Color Appearance” tool and the “Viewing Conditions” settings are “dangerous” in that regard. Hence the questions.

Well, I’m supposed to be at work in seven hours and I still need to shower and stuff. And get some sleep, obviously. See you, everyone. Thanks to anyone willing to shed some light on any of this.

[priort (Todd Prior)]

It sounds maybe like your color management is broken…its likely consistent as it sounds like its being treated as srgb or a default profile that some of your other software is using??

All that is a bit of a moving target though …

Aside from that, I don’t think color appearance would have anything to do with it…you process your edited image to the selected output profile and that is what should be used by color managed applications…The viewer has no idea what modules were used its just pixel data in the defined color space…

I’ll have to read all that again …as for web browsers there are several test pages to assess color management …it might be good to try that and also maybe some of the test icc profiles that can confirm your software is behaving as well…

[alicem]

Thanks for your answer here and PM. I will try to respond to both here for the sake of sharing information.

@priort

It’s AI search but it has the answer to why the setting is set to false.

Haha, I know this site isn’t fond of IA stuff, but I think your LLM-generated info could have been placed in a “hidden details” tag so that anyone interested could read it.

I see that:

The gfx.color_management.mode setting in Firefox controls how color management is applied. The values are: 0 for no color management, 1 for full color management (applied to all content), and 2 for color management only on images with ICC profiles.

(The about:config could give some built-in explanation for cryptic integers like that… Are we supposed to run a web search for every item? )

• Mine is 2 and it says it’s the default, as surmised by the text you pasted.
• People seem to say that changing that can fix our JPGs but ruin all the rest.
• No gfx.color_management.display_profile at all. I never fiddled with that, after all, so the opposite would have surprised me.
• gfx.color_management.native_srgb is already false (the default).
• My Firefox is up-to-date (or at least it was yesterday).

Actually, I read yesterday that even basic CSS-produced colors can look different across browsers depending on settings and stuff, not only images! That’s horrible. Glad I’m not a web designer.

My monitor is relatively cheap: I went to a store in which it was the only IPS screen available (I had been told that this technology was more suitable for photography, and I indeed get the impression that it has OK-ish colors out-of-the-box). But perhaps even cheap IPS has a wider gamut? Will need to do some tests on the laptop’s screen itself, but preferably after a boot during which the external screen is not plugged in at all, to make kinda-sure there is no weird interference.

@priort

I don’t think color appearance would have anything to do with it

Well, in my blooming paranoia and piled-up exhaustion (gosh, why do I use such literate-sounding phrases in English? I’d never talk like that in French LOL), I started imagining that there could be strange metadata fields crammed into the exported image, saying stuff like:

Hey, by the way, when displaying this picture, try to do this and that in order to adapt to the viewing conditions! The aforementioned viewing conditions were described as being blah blah by whoever exported that picture! Thanks!

… a metadata field that could have been honored or ignored depending on the used app and app settings. I did not seem too far-fetched, but you’re probably right in saying that there is no such thing.

Well, your input seems to hint toward mere weirdness either on my machine or in Firefox, and seems to indicate that I do not need to loop through my whole picture series again to adjust the .pp3-s, am I right? Especially with Chrome being sadly used by nearly everyone these days… And it’s not like the Firefox version looks ugly either – it’s probably way easier for me to spot that something is amiss because I’m the author.

Edit

Still, there’s the question of “wouldn’t it be more robust to upload, on the web, pre-converted, profile-less pics?”.

[syyrmb]

I’m not an expert in Color Management, perhaps not even a noob, but from what I understand, color management in general is a difficult (and expensive) thing to do right. You need a color calibrated screen before talking about if a picture shows the right color or not.

I’ll provide my takes on your questions, but they might be very wrong, so hopefully someone can correct me before OP sees them.

Q1. Re: Firefox’s color management option: I read from an article from Chinese’s Quora-ripoff website saying Firefox is the only browser that can actually respect the picture’s color profile and display the correct color, if you do it correctly. But if you allow OS like Windows (or probably MacOS) to take over color management, you could be screwed, especially if the image is in wide gamut.

Q2 & 3. Re: CIECAM “viewing condition”: My understanding is that CIECAM is a tool to take photographing scene’s conditions, post-processor’s environments and viewer’s viewing environments into account and make compensation to ensure that the post-processor and viewer sees more or less the same image. I believe this only work when yours and the viewer’s screens are calibrated. It’s part of the color management workflow but you do need a calibrated screen to make use of it (or you can just use it to achieve some fancy effects).

Q4. Re: Choosing color profile for consistency and printing: I believe you need to know the print shop’s printer’s color gamut and use that to do soft proofing in Rawtherapee to get a relatively accurate result for your printed work (and we haven’t even taking the printing paper’s “gamut” and dynamic rage into account). If you’re talking about online viewing in browser, your best bet is probably use an srgb monitor, output image with srgb color space profile (or no icc profile, which defaults to srgb) in RT so that everything is under srgb no matter what. Pros use wider gamut which I don’t mess with so can’t say anything about it.

Q5. Re: RTv4_SRGB and how profile works: I personally uses the srgb profile with that longer name. I think RT embeds it into the picture since when I open a picture in GIMP, GIMP asks me whether to use RT’s profile or switch to GIMP’s default profile.

Q6. Re: Why embed a profile rather than hard-code RGB value: I guess it probably has something to do with color management. How I understand it is that, if a flower has large area of RGB (255, 0, 0) with no profile at all, then it’ll display the reddest red of your monitor, and that’s gonna be very unpleasant to look at if you have a wide gamut one because it would be too red. a profile would tell somebody to convert this “255” red from (for example) sRGB to the relative red value in (something like) aRGB to keep consistency.

So, just use sRGB seems to be the most worry-free method.

[alicem]

@syyrmb

hopefully someone can correct me before OP sees them

That’s very thoughtful of you I will check back in case you edit stuff.

@syyrmb

from Chinese’s Quora-ripoff website

@syyrmb

Firefox is the only browser that can actually respect the picture’s color profile and display the correct color, if you do it correctly

I suspect Firefox is trying to do something intelligent like that and kinda fails in my setup – I dunno, but that’s gradually becoming my main hypothesis.

@syyrmb

CIECAM “viewing condition”: […] you can just use it to achieve some fancy effects.

So it’s not something “dangerous” like when white balance screws up skin tones when (mis)used for artistic purposes?

@syyrmb

do soft proofing in Rawtherapee

That could be a whole dedicated topic I’m sure, but I don’t know at this point whether soft proofing in RT is “RT automatically adjusts to the given printer’s profile by computing some sorts of deltas, and the exported image looks different despite the .pp3 being the same” or “it alters the preview so that you can fiddle with random settings until it looks satisfactory and produce a separate image with a different .pp3, specifically for printing”. But I could probably find out by looking for info. Just not my top priority at all right now. xD

@syyrmb

If you’re talking about online viewing in browser, your best bet is probably use an srgb monitor […]

Do you think that the “Color space: rgb” mentions in my Xubuntu settings screenshots from my “Expandable details about my system and my very basic, very default-ish settings” hidden details block in the initial post proves that I’m using sRGB monitors only at the moment?

@syyrmb

… output image with srgb color space profile (or no icc profile, which defaults to srgb) in RT so that everything is under srgb no matter what.

And I suppose RT’s RTv4_sRGB is OK, then?

@syyrmb

I think RT embeds it into the picture since when I open a picture in GIMP, GIMP asks me whether to use RT’s profile or switch to GIMP’s default profile.

Yes, I noticed this as well. This is how I produced a profile-less version, by importing and then exporting (while explicitly unchecking the profile box in the export options).

@syyrmb

it’ll display the reddest red of your monitor, and that’s gonna be very unpleasant to look at if you have a wide gamut one because it would be too red

Oh, I did not think of such possibilities. So uploading profile-less pictures on the web could lead to ugly results for people with such monitors, and should therefore be avoided?

Meanwhile, I tried looking at my pictures again. On a brightly lit monitor (and after some rest…), the difference is a bit less obvious – and probably even less so for people that are not the author. In some cases, I even think Firefox’s version looks more pleasant, but I guess humans have a bias toward warm colors (I try to avoid over-relying on this bias while processing, but sometimes fall in that trap nonetheless xD).

Since I already had the intention of “advertising” this picture series in the “Showcase” section of this website (even if this kinda scares me), I think I’ll add a note asking people whether anything looks prohibitively “off” color-wise in their personal setups. Well, one risk is that over-sensitive persons and color-savvy pros might point out imperfections that have nothing to do with color profiles and more with my actual, intended processing choices, but hopefully this will be reassuring overall.

In the meantime, here’s a ZIP (to avoid any server-side black-box processing) containing the HQ (well… relatively to what I previously posted, I mean) versions of a same picture: the original RT export, and the GIMP-produced profile-less version:
with-and-without-profile.zip (15.7 MB)
In my Ristretto image viewer, in GIMP, and in Chromium, they look 99–100 % alike to me, while Firefox shows the aforementioned differences (but still looks pleasant if you don’t know any better or even if you don’t try to compare and let your brain adjust). I did not try zooming in like a brute, but still.

I would be interested in:

• What people get on their side. Edit: Especially on Windows, as I haven’t used Windows for about ten years.
• If it looks “good enough for a non-pro whose first non-compact camera is not two-year-old yet regardless of the profile” to you.
• In case you see a difference: whether one looks terribly odd or if both look good anyway.
• Edit: In case you don’t see a difference: whether it looks more like the red-ish or cyan-ish version from my .webm in my initial post (in the last expandable hidden block).
• (For hardcore people) If you see anything strange that could be related to that stuff in the metadata.
• If you have a non-sRGB weird high-end fancy monitor, it might be interesting to see if the profile-less version goes to hell, and to what extent (assuming hell is subject to gradation ).

Since I know for a fact that some people just send phone-issued JPGs to print shops without even knowing that such a thing as color profiles even exist and are happy with whatever they get, it can’t be that hard to get something nice enough for the average viewer, can it?

[syyrmb]

@alicem

@syyrmb

CIECAM “viewing condition”: […] you can just use it to achieve some fancy effects.

So it’s not something “dangerous” like when white balance screws up skin tones when (mis)used for artistic purposes?

I think it’s just another set of pixel transformation tool tailored for color appearance usage, how it behaves depends on the user’s intention and execution. It does have a WB option, so maybe it can cause weird off-color if used without discretion.

@alicem

@syyrmb

If you’re talking about online viewing in browser, your best bet is probably use an srgb monitor […]

Do you think that the “Color space: rgb” mentions in my Xubuntu settings screenshots from my “Expandable details about my system and my very basic, very default-ish settings” hidden details block in the initial post proves that I’m using sRGB monitors only at the moment?

Not really because I didn’t look at any screenshots. But if you do have a wide-gamut monitor, it could still be set to sRGB mode, which (I believe) is suitable for most people.

@alicem

@syyrmb

… output image with srgb color space profile (or no icc profile, which defaults to srgb) in RT so that everything is under srgb no matter what.

And I suppose RT’s RTv4_sRGB is OK, then?

Rawpedia says it’s close to sRGB. I personally use the real sRGB just to be safe.

@alicem

I would be interested in:

• What people get on their side. Edit: Especially on Windows, as I haven’t used Windows for about ten years.

I’m using Windows and I see two visually identical images.

@alicem

• If it looks “good enough for a non-pro whose first non-compact camera is not two-year-old yet regardless of the profile” to you.
• In case you see a difference: whether one looks terribly odd or if both look good anyway.
• Edit: In case you don’t see a difference: whether it looks more like the red-ish or cyan-ish version from my .webm in my initial post (in the last expandable hidden block).

It looks ok to me. I was gonna say the dried grass is a bit reddish to my taste (I tried to tone it down in RT), but your .webp is less reddish and closer to my adjustment in RT (still a bit redder to it though).

@alicem

• (For hardcore people) If you see anything strange that could be related to that stuff in the metadata.

No strange things to me in general, but the profiled picture is smaller than the non-profiled one. A bit curious.

@alicem

• If you have a non-sRGB weird high-end fancy monitor, it might be interesting to see if the profile-less version goes to hell, and to what extent (assuming hell is subject to gradation ).

I don’t follow own advice and am using aRGB mode without any color calibrations (technically a very bad idea, but fortunately I don’t rely on color accuracy to make a living). The picture comes out right.

@alicem

Since I know for a fact that some people just send phone-issued JPGs to print shops without even knowing that such a thing as color profiles even exist and are happy with whatever they get, it can’t be that hard to get something nice enough for the average viewer, can it?

Not color calibrating is not the end of world. Many Android smartphone’s screens are somewhat calibrated out-of-factory (at least in China). But if someone’s screen is cyan or too yellow (like some of my colleague’s ridiculous monitors), that could be a problem.

Your other questions/comments are beyond my current scope and I can’t provide any insight. Hopefully my non-native level of English and less-than-ideal technical insight aren’t a bother to you.

[priort (Todd Prior)]

Can it be that firefox needs to see the profile explicitly and lost in the text above maybe that is how you are set up…so Firefox is set to color manage but the profile you use is not seen by it so it treats it as untagged…

I’m not sure…but if you have the means via the firefox setting to specify everything ie the display profile and place the RT output profile in the OS folder that stores ICC files maybe it will work as intended…

Just thinking along the lines of the reply from @syyrmb Q1 comments…

So following on here you could be sure that you use the v2 variant of the profile and specify implicitly your display profile to FF and then experiment with CM options 0 1 and 2 to try to see by the process of elimination how FF is handling the CM for the image…

ICC color correction in Firefox - Mozilla | MDN

Thats for a bit of an older version but the overall info might still be correct??

[alicem]

@syyrmb

It does have a WB option

Is that the “Chromatic adaptation” slider I mentioned, or something else entirely? Even when setting the Color Appearance tool to “Advanced” mode, I only see something called “Gray balance (Slope)” – nothing explicitly labeled as a white balance setting.

@syyrmb

I’m using Windows and I see two visually identical images.

Thanks. May I ask which app(s) you tried?

@syyrmb

I personally use the real sRGB just to be safe.

That would be interesting to try, to see if it alleviates the per-app differences. But with my current knowledge I am not able to say how big of an impact it would have on the result overall – as in, whether it would invalidate my work and force me to adjust stuff drastically. Perhaps I’ll try for future edits rather than fiddling with old ones. I think RT’s RGB had an advantage, though – I should go back to RawPedia some day to re-read about it…

@syyrmb

the dried grass is a bit reddish to my taste

I may have exaggerated on some parameters. Vibrance 5 / 12, a bit of color toning… The RGB curves, strangely enough, has negative values for each range of red (in parametric mode). And for that picture at least, I did not do anything weird with the “Chromatic adaptation” of color appearance.

@syyrmb

No strange things to me in general, but the profiled picture is smaller than the non-profiled one. A bit curious.

You mean in terms of file weight? Don’t overthink this – I think I did something stupid like using a higher quality level for the GIMP export than I did for the RT export, so GIMP probably strove hard to accurately encode a bunch of artifacts, haha… Sorry about that.

@syyrmb

too yellow (like some of my colleague’s ridiculous monitors)

Maybe just a blue-light filter to fight eye fatigue? I sometimes enable this pretty aggressively, and my colleagues are like “What the hell?!”. But of course I turn it off when using RawTherapee (RIP my eyes and head when bingeing like I did those past few weeks).

@syyrmb

Your other questions/comments are beyond my current scope and I can’t provide any insight.

Thanks for what you did anyway, especially considering that it may provide a basis on which others might add details later.

@priort

if you have the means via the firefox setting to specify everything ie the display profile and place the RT output profile in the OS folder that stores ICC files maybe it will work as intended

Maybe, but:

• It seems simpler (at least in my case, perhaps not in general) to just set the weird about:config boolean to true.
• I can’t force the (arguably few, alas) visitors of my website to do this. I personally don’t care much if my Firefox shows absurd things, as long as:
  1. we are sure that it’s not every other app that is messed up and Firefox the least-broken one (like in many domains, the largest crowd is not necessarily right),
  2. things look at least OK-ish for other people on their machines on average, and
  3. I have other means of viewing my work, including a hopefully reliable RT preview.

Maybe it’s a case of “Firefox is different (when using the default settings) and we need to bite the bullet and accept minor deviations”…

Edit: There are tons of open bugs for an sRGB search (for images, CSS stuff, and even videos!), many still open and some more than a decade old. It took them ages to implement support for v4 profiles, too. https://bugzilla.mozilla.org/buglist.cgi?quicksearch=sRGB

I was asking about printing, but maybe the differences between printing services are even worse than that? I always assumed that soft-proofing only needed, like, a slight bump of the “L*a*b* Adjustments” sliders this way or that way, but maybe it’s far more complex than that. That would make me relativise this whole Firefox business…

[nosle]

As others have mentioned your problem seems related to colour management and not you choice of processing.

There are basically three components

1. How your os colour managment is configured
2. How your app colour management is configured
3. How RT is exporting and tagging your file

You should also be aware that the people viewing your images online may have any sort of crazy settings. Extremely low or hight contrast, huge colour casts etc.

If you havent characterized your display the minor difference between Firefox and RT may actually be overshadowed by you editing with “wrong” colours. It’s normally ok though but if you’re very precise about your edits you kind of have to calibrate/characterize your display.

[syyrmb]

@alicem

@syyrmb

I’m using Windows and I see two visually identical images.

Thanks. May I ask which app(s) you tried?

Sorry, I intend to include that info but forgot to do so. I used some random image viewer in Bandizip, Rawtherapee, GIMP, and also a Firefox-based browser. All the same result.

@alicem

@syyrmb

too yellow (like some of my colleague’s ridiculous monitors)

Maybe just a blue-light filter to fight eye fatigue? I sometimes enable this pretty aggressively, and my colleagues are like “What the hell?!”. But of course I turn it off when using RawTherapee (RIP my eyes and head when bingeing like I did those past few weeks).

It’s some generic dell/benq monitors. Most of them have color biased to magenta (for some reason I said cyan) and/or yellow. We use 2 monitors at work, so everyone (except me)'s monitors have different color. How anybody can tolerate using 2 monitors that have different and very wrong colors side-by-side is beyond me, but apparently nobody seems to think that is a problem which shows color management isn’t an average user’s concern, pretty much.

[alicem]

@nosle

There are basically three components

1. How your os colour managment is configured
2. How your app colour management is configured
3. How RT is exporting and tagging your file

Thanks. Well, everything is using the default settings (as far as I know). I don’t think the colors are terribly off because I like what I see when I display the result on most other devices and stuff.

@nosle

the people viewing your images online may have any sort of crazy settings

Hehe, yeah, I just hope my friends and family are respectful enough to at least try to check my stuff in acceptable conditions.

@nosle

if you’re very precise about your edits you kind of have to calibrate/characterize your display

Yeah, I might do that some day, but:

• I think I would then be even more worried because editing while thinking you have properly applied profiles when you actually haven’t done so correctly could probably be worse, so it’d generate some stress for me.
• As I mentioned, this is not a topic that I am fond of at all, so I tend to try to push it away from that hobby which is supposed to help me relax.

And seriously, that huge list of sRGB-related bugs collecting dust on Bugzilla (Edit: with some explicitly mentioning oversaturated reds, which is what I see) hints at Firefox just being a bit lame in how it handles RTv4_sRGB. Will need to see how it behaves with a basic sRGB pic instead.

@syyrmb

I used some random image viewer in Bandizip, Rawtherapee, GIMP, and also a Firefox-based browser. All the same result.

Very cool, thanks! It’s probably more representative of the average population than my tests on Xubuntu… And GIMP and RT provide some common ground between our setups.

@syyrmb

How anybody can tolerate using 2 monitors that have different and very wrong colors side-by-side is beyond me, but apparently nobody seems to think that is a problem which shows color management isn’t an average user’s concern, pretty much.

I honestly don’t care when I’m at the office, but it depends on what you do, and on how often your gaze swings from one monitor to another. Having my music on one screen and a source code editor on another, for example, makes botched colors a non-problem in my book.

[syyrmb]

@alicem

@syyrmb

It does have a WB option

Is that the “Chromatic adaptation” slider I mentioned, or something else entirely? Even when setting the Color Appearance tool to “Advanced” mode, I only see something called “Gray balance (Slope)” – nothing explicitly labeled as a white balance setting.

I forgot you were using CIECAM in the local adjustments tab. The CIECAM tool in Advanced tab has the complete feature set and is what I’m more familiar with, so that one is what came to my mind at first when you were talking about the CAM stuff.

@alicem

@syyrmb

How anybody can tolerate using 2 monitors that have different and very wrong colors side-by-side is beyond me, but apparently nobody seems to think that is a problem which shows color management isn’t an average user’s concern, pretty much.

I honestly don’t care when I’m at the office, but it depends on what you do, and on how often your gaze swings from one monitor to another. Having my music on one screen and a source code editor on another, for example, makes botched colors a non-problem in my book.

Maybe because my job requires me to switch between the 2 screens back and forth, or maybe I have a bit of OCD, I don’t know yet.

[alicem]

Here I go again (instead of taking some rest):

Older machine…

… with Xubuntu 22 instead of 24, and possibly a cleaner install.

I booted up an old laptop and downloaded my own ZIP from before. The two pics look the same, even in Firefox, and they match the “colder” version from RT.

A major difference between the two laptops is that the newer one has that weird “Automatic - Built-in display” color profile in the OS settings for the internal monitor (see the screenshot from initial post), while the older laptop has none. Simply toggling this off makes Firefox agree with other apps again, despite the fact that I’m on the external screen.

Edit: By the way, I got the current machine around February, which could very well explain why I did not notice any difference for my older pictures.

Well, now I feel like an idiot because, contrary to appearances, this was the last section I wrote and the rest of the message seems a bit overkill. I’ll still post it because it may provide interesting discussion points…

Different output profiles in RT

I tried producing a JPG with RTv2_sRGB instead of RTv4_sRGB set as the “output profile” in RT, among other tests.

• The good news is that it does not produce visible changes in the output, so it does not invalidate my old edits.
• The bad news is that the only JPG for which Firefox agrees with the RT preview and other apps is the one produced with “No ICM (sRGB Output)” as the output profile – which apparently means that there’s no profile at all, as GIMP says nothing about converting any profile when opening that particular version. And as I gathered from our previous posts in that thread, uploading profile-less pics on the web is far from ideal, right?

For a randomly-chosen already-available-in-RT profile, RTv4_Rec2020:

• The colors are very subdued in my image viewer…
• … but I don’t see any increase or decrease in Firefox’s warmness, as if Firefox simply processes all embedded profiles in the same (possibly wrong) way. Or stacks them on top of whatever weird thing it does.

Am I right to assume that a “sane” application should produce a different output when switching profiles like that (assuming the considered profiles are different enough, which seems to be the case between the RTv4 flavors of sRGB and Rec2020), and that the fact that Firefox sticks to the same result makes it the “anormal” app in the bunch?

It seems that output profiles in RT, unlike working profiles, do not alter the preview – only the way the exported images will render when eventually opened in other apps. This can probably lead to bad surprises. I guess the smart way to use those is to select the same profile as the working one?

“Rendering intent” in Firefox

I tried all possible values for gfx.color_management.rendering_intent, with the JPGs exported from RT with different profiles as explained above. Did not see any change: still warmer. I even got the impression that, at least for my photos, that setting was being completely ignored (which might be normal, I dunno).

Non-RT profiles

I got a standard-as-hell, non-RT profile from color dot org (thanks @priort) and stored it in the directory that RT mentions in its settings:

sudo cp -vn ~/Downloads/sRGB_v4_ICC_preference.icc --target-directory=/usr/share/color/icc/

… but I don’t see it in the working profile and output profile dropdown menu lists in RT, even after exiting and relaunching RT. The settings and permissions look right to me: Every user can read the profile and access the parent directories:

  $ ls -l /usr/share/color/icc/
total 68
drwxr-xr-x 2 root root  4096 Aug 27  2024 colord
drwxr-xr-x 2 root root  4096 Jul  8 19:01 ghostscript
-rw-r--r-- 1 root root 60960 Aug 29 18:33 sRGB_v4_ICC_preference.icc

  $ ls -l /usr/share/color/ | grep icc
drwxr-xr-x 4 root root 4096 Aug 29 18:33 icc

• Yeah, there were two subdirectories there already. Dunno if RT expects me to create one for RT with a specific name.
• Interestingly enough, the existing subdirectories contain tons of profiles already, including one simply called sRGB, but RT does not seem to search that place recursively and only lists its own built-in profiles.

Not sure what I’m doing wrong.

GDK, the default monitor, and working profiles

If I check the following box in RT (which was not the case previously):

… the working profile gets sets to “System default (from GDK)” instead of the “None” I’m used to. (I can also switch directly at any time via the dropdown menu at the bottom of the RT editor.) And… this yields a very, very Firefox-ish output!

I guess that’s some progress in our investigations? That brings yet other thoughts, though:

1. If that’s the “System default”, why would no app except Firefox use it?
2. What’s that stuff, in the RT settings, about this being the profile for my default monitor? As far as I can see, the colors look extremely close between my default (laptop) monitor and my external one…
   • NB: I generally use RT on the external monitor – even if I did recently process some pics on the laptop directly, noticing that it yielded basically the same result anyway (it’s just not very ergonomic to use RT with just the laptop), which is one more reason to be surprised with this stuff about the default monitor’s profile making everything distinctly warmer when summoned.
3. If the Firefox setting native_srgb is supposed to let the system handle the colors when it is set to true, why does it look like Firefox does the exact opposite (GDK-like colors when false, and agreeing with other apps when true)?
4. Why does Firefox seems to use the GDK-like colors only for JPGs containing a profile, and refrains from using it when there’s none in the JPG? This, too, sounds like the opposite of what I would have expected.

But frankly, if (and only if?) the answer to my…

Am I right to assume that a “sane” application should produce a different output when switching profiles […], and that the fact that Firefox sticks to the same result makes it the “anormal” app in the bunch?

… is something along the lines of “Yes, exactly: your Firefox is drunk; ignore it and have faith in RT’s preview”, Firefox’s color management can go to hell and I’d just leave things like that.

PS 1: Funny how I manage to gradually learn stuff – without too much pain apart from the fact that I’m getting more tired than ever – about this dreaded topic thanks to your help. This is much appreciated.

PS 2: See the first section of this message for the biggest “breakthrough”. I think it is now safe to say that my pics are OK and that I do not need to change anything in their .pp3.

[cedric (Ted Cousins)]

@syyrmb

@alicem

@syyrmb

… output image with srgb color space profile (or no icc profile, which defaults to srgb) in RT so that everything is under srgb no matter what.

And I suppose RT’s RTv4_sRGB is OK, then?

Rawpedia says it’s close to sRGB. I personally use the real sRGB just to be safe.

“v4” is ICC’s later profile version which goes beyond their previous V2 in terms of options. See:

Why use ICC v4 profiles?

[alicem]

@cedric

“v4” is ICC’s later profile version which goes beyond their previous V2 in terms of options.

The interrogation was more focused on how well the RT flavor of it is supported by apps around the world and how different it is from a non-RT-prefixed flavor. But well, it seems that in my case the exact profile was not to blame (even if the technical details somewhat escape me).

[cedric (Ted Cousins)]

You can test your system for V4 compatibility here:

Is your system ICC Version 4 ready?

[alicem]

Interesting. Support seems OK (I guess nearly all systems nowadays support v4?). But yeah, even with their image, I get a different result in Firefox depending on whether that weird “Automatic - Built-in display” color profile in the OS settings is enabled (I opened the test each time in a new private browsing session to avoid caching and stuff). I think I’ll keep it disabled. I don’t even see the impact on my laptop’s screen display: literally the only effect that I see is that it messes up with pics in Firefox.

[cedric (Ted Cousins)]

@alicem

Interesting. Support seems OK (I guess nearly all systems nowadays support v4?).

Glad to be of help!

[alicem]

I added a TL; DR in the initial post (I think that if I wait too long it’ll become uneditable ). Maybe this could help others.

[priort (Todd Prior)]

@alicem

Not sure what I’m doing wrong.

C:\Program Files\RawTherapee\5.11\iccprofiles\output

Is there the equivalent subdirectory in Linux…maybe try there…