So here is the deal the display profile comes in before the histogram profile. My screen is dark because I set the display profile just for the gamut check to linear rec2020 … I think for darktable the srgb profile will work also because it is unbounded… But to remove this possibiltiy at all you set working, display, and histogram to linear rec 2020… now when you check the overexposure set to full gamut you can see what in the pipeline is gamut clipped… When you select the gamut clipping tool that uses what you have set for softproofing… if its srgb then what it shows is all the gamut that won’t fit in the space…but this is handled by color management and mapped…otherwise if you worry about this then you might as well set it to srgb for everything and that would not make sense… So in a way the gamut warning shows you what will have to be mapped and the overexposure gamut will show you what cannot be accurately mapped to your desired output… I hope that makes sense…
EDIT
So in my mind and I could be wrong but set-up as described above with your softproof profile matching your export or output and your histogram profile matching your working profile and then confirming that your display profile is well behaved by checking what impact changing it to linear rec2020 has… it should be none if it is then that display profile distorts true gamut readings… So then from this diagram A would be rec2020 and B would be SRGB. The gamut warning tool would show the areas shown here as out of gamut and needing to be mapped as well as any area’s beyond that and the overexpose gamut would show you any data clipped outside of the A color space ie rec2020 and so not able to be correctly mapped by the colorspace conversion during export… for sure others here are far more knowledgeable than I but this is how I understand it to work and I think AP has decided to remove the histogram profile in his fork and try to tackle this issue so the display profile cannot be an issue ever…
It’s a Sony file with 4.2, that gets it from the maker notes, so it’s alright.
Still, in highlight-problematic files I often do a double-check anyway (to find magenta areas of blown parts, and if they are neatly indicated as clipped).
It is worth checking for sure. One recent playraw had 16000 or more and the exif was like 12500 or something for specular… I think it was a Canon file… loose recollection but a big difference in how all the math would work that relies on the raw white point…
Exposure set and then exported from darktable as unbounded, 32-bit linear TIF. I did not touch contrast, so it is low. There was some compression applied to highlights; colours were mapped to sRGB. Both tone compression and gamut compression were done in OkLAB, using my toy project.
@priort , I think I understand relative.col vs. perceptual output and indeed I generally use a profile that support perceptual, the “ICC Preference perceptual” which I’ve mentioned previously. It’s the jiggery pokery in DT I don’t get at present, need to invest some time.
Here are the peppers using @jorismak 's XMP with sRGB websafe and then the pref. percep one. I think the latter looks rather better.
According to Aurelien perceptual and relative intents don’t work in DT… I think you need lut based profiles for these to work but I am not sure… I just presented that diagram to emphasize the gamut being displayed by the two different indicators in DT…
EDIT:
I was on my phone before but I can see the difference now on my PC so something appears to be working here by choosing different intents…
EDIT2…
I can’t replicate what you were able to produce
With my xmp or the one you used and all combinations of settings LittleCMS off or on… setting intent with output profile and leaving export to image settings or at the export or setting both the to match I get the identical image with no observable difference between relative and perceptual. I wonder what others would see trying the same…
Edit3
Looking at the exif data my jpg are both rendered as perceptual despite having different settings applied… I wonder if my OS is doing something here…
But looking back I think I am seeing what others have so maybe I misunderstand what profiles you are using because I dont think the websafe one in DT will support intents… Colour management - intents
The pepper needed some TLC.- two instances of color balance RGB to color grade two different casts in the bright areas at the top.
Toned down the background and left/right sides a bit. Not a vignette, but just to subtly move attention to subject.
Modified filmic v6 used, with 0.35 threshold on highlight reconstruction.
Guided Laplacian to tame the reflections on the pepper a tiny bit, though it is barely noticeable.
Tone equalizer to bring down some overly bright highlights.
Blur using diffuse/sharpen “simulate watercolor” in subtract blend mode to bring out some surface details. Then add a small exposure boost to bring the luminance back to the original level.
What I see for this image is that all 3 look identical when viewed in krita (3 looks slightly different to 1 and 2 in geeqie and firefox, perhaps because those apps don’t handle v4 properly?) . None of them look like your perceptual output.
@Soupy , nothing changed in my XMPs (except output profile)! Presumably you said to use little CMS in DT settings? And perhaps check you put them in the right folder. sRGB_v4_ICC_preference.icc is the right one from the page you linked to. Have a look at your Terminal log, I get an error message using this profile, “unsupported profile” or something like that. If you don’t get that, you are probably not using it for some reason, wrong folder I’d imagine. AurelianP once said something about this, I think the message arises because the profile is incompatible with parts of DT but nevertheless goes thru to LCMS ok.
[dt_ioppr_set_pipe_output_profile_info] unsupported output profile 0 /home/usr/.config/darktable/color/out/sRGB_v4_ICC_preference.icc, it will be replaced with sRGB
Which folder do I move the profile to avoid the error? Or which setting do I change?
Don’t worry about that message. It will be applied as the output profile. See here:
[the error message] just tells that darktable won’t be able to use profile internally for some things […]
However, the output color profile module will apply the requested color profile and intent (using Little CMS 2) at export.
I don’t know if it matters , but i had a sRGB icc file selected in the output profile module i guess.
I use Elle’s v4 profiles and export to linear (g1). Do my downsizing in an external app and that converts to regular sRGB trc.
That’s also where i export to rec2020 linear , with Elle’s profiles … but since filmic can’t give a good preview of the rec2020 output , i try sRGB more these days.
(Still frustrated by the lack of filmic softproofing… might actually dive into the code myself ).
Thanks @kofa
I tried again with both v2 and v4 perceptual profiles, compared to websafe.
I made sure to compress history before each export (in case that affected filmic gamut mapping, not sure).
v4 perceptual was identical with lcms2 on or off.
v4 perceptual was very similar to websafe relative, but noticeably different.
v2 perceptual was near identical with lcms2 on or off, difference hardly noticeable.
v2 perceptual was near identical to websafe relative, difference hardly noticeable.
This shows that the different profiles, intents and lcms2 settings are actually doing something, which is good. However, none of it looked like the perceptual result @RawConvert posted above, so I’ve no idea how he achieved that.
I do too, but want to compare / fine tune with perceptual profiles.
