New Primaries module

Testing now :heart_eyes:

5 Likes

Me too! Just got the latest nightly windows build. Looking good.


Just to confirm, am I right in thinking that anything you can do with this is already possible with the channel mixer in color calibration, just less straightforward? (significantly less depending…).

2 Likes

Yes, you’re right. Channel mixing in color calibration is even more generic i.e. more degrees of freedom.

4 Likes

Thanks! I like this new module though - seems much more approachable then the full blown channel mixer. (in my case though I’d just got the hang of the latter - now I can start over again :smile:)

Edit. Also, is this similar to the Adobe ‘camera calibration’ approach? I think so but I’m not familiar with that one. Another good point for anyone migrating away from Adobe if so.

1 Like

Was about to say: looks filmic (and not in the filmic rgb sense), but then I saw the Green Film Look rgb curve enabled :smiley:

2 Likes

11 posts were split to a new topic: the need for a scene referred color zones replacement

A few days ago, trying to get my head around the channel mixer,
(I never managed it), I asked here for something simpler. It seems my wish was granted.
Thank you.

4 Likes

I have a very stupid question… so when ignoring the soft limits and pushing the hue slider at some point the display jumps to black then white I think. Is this the math or is this a display preview thing… I tried to think of what I was seeing but it was early this morning and I was playing with a couple of colorwheels and and NEF colorchart file…

Just stupid but curious??? Thx :slight_smile:

EDIT. I just downloaded the nightly and tested on my work PC…different display driver here… The response was more clear cut. I think when the setting drives the pixels out of gamut it goes black. It does go black and then come back and then go black again as you keep scrolling…these are extreme settings mind you ie greater than 100 degree rotations… but the interesting thing was my PC at home showed a different pattern of the vectorscope so I am thinking that PC which has a calibrated monitor has a profile that is interacting the data that is going to the histogram profile and hence what I see on the vector scope. I will try to do at test across the range of the sliders on both PC and with the profile set to srgb and rec2020 to see how they behave…

Could you please provide an XMP that reproduces the black issue? Specifics of the two systems would be also helpful (especially which GPU / driver).

Ya I will attempt to do something better than my casual observations

@flannelhead and all the devellopers team
Congratulations on this module. I really like its ease of use, I can use it without reading the manual.
Many thanks, A simple and efficient module :clap: :clap: :clap:
Christian

5 Likes

Here is the xmp… around -120 or +120 as you use the scroll wheel it will jump to black and then back the next rotation of the scroll wheel. If you move the purity slider one tick the color comes back… Its extreme and not something you would likely do as things seem out of gamut at that point but maybe it points a small math thing… also when you right-click the slider and select a negative value it jumps to the positive one initially and the reverse…

XT30hSLI00100NR1.RAF.xmp (20.3 KB)

Window 10 for this one and older 745 Nvidia card but my machine at home was much newer and I think I saw something similar… just quick messing around… as I said I will try to be systematic at home later and maybe its just a case of pushing things to weird extremes…

Is that Bill’s weekly or another build?

Thanks.

1 Like

Not Bills, got it here: Releases · darktable-org/darktable · GitHub
I expect it’s obvious to people who know github but I was unaware of it until someone kindly showed me!

1 Like

Thank you for the example. I think I have a clue why this happens, but will have to analyze deeper and think if a safeguard can be added without crippling legitimate use cases. As far as my current understanding goes, things like this can happen when the whitepoint no longer lies within the triangle formed by the primaries in the xy chromaticity diagram.

Don’t know what Lightroom’s “camera calibration” does under the hood, but definitely the interface seems similar. I’d guess it’s also fairly similar processing.

3 Likes

I literally was just messing around this morning for 10 min while I had a quick coffee so I thought I would pull up the vectorscope and mark a patch on the image and then start watching how it moved and how the other colors were impacted when I moved sliders… On my PC at home the yellow patch moved as if rotating and then circled in towards the center. I didn’t seem to drive it out of bounds like I did here on my pc at work… so I need to go back and check… That is why I was wondering if my display profile was somehow getting in the way… I will also check that…

1 Like

Finally played around with this today. A nice addition, fun module! Certainly easier to grasp than channel mixer. Worth to compare ‘normal’ blend mode with ‘chromaticity’.

I’m wondering if it might be better to increase the soft limits from 20 to 45? 20 allows a change from blue to teal, whereas 45 allows a change from blue to cyan. For achieving results close to reality 20 is good, but more creatively, to achieve certain harmonies when colours need to be pushed a little further, 45 would help. I haven’t played around enough to know what a standard adjustment would be, and it no doubt depends upon subject matter and taste, its just that on the first picture I played with I needed to set one of the hues to 40 to achieve a harmony.

1 Like

Generally yes, because it can adjust saturation etc, but if it is only about the RGB matrix part (channel mixing proper), then my understanding is that the two modules have exactly the same degrees of freedom.

The reasoning is this: the primaries module has 8 sliders, while the 3x3 RGB matrix has 9 entries, but if we consider exposure a normalization (eg that one would not use an c\cdot I_3 matrix to adjust exposure and exposure only), then it has 8 degrees of freedom (+ exposure).

Hello,

Looking at the photos of the Sigmoid project https://github.com/darktable-org/darktable/pull/15104 , I can’t wait to test it and see what happens :wink:
Will the “primaries” also make it possible to recover details in the highlights with the new version of Sigmoid ?
Greetings from Brussels,
Christian

I think for the most part this is a color tool but you might be able to tweak it in such a way to give a perceptual bump I don’t think your making a big correction to offset the detail loss from compression