Thanks @anon41087856, I like the first two, especially the first. I see how you’ve used “global” alone for the de-saturation.
I think I’ll look at using the purity sliders fairly routinely, they seem to give a pleasing effect even if you don’t particularly want a pastel look.
@priort
I used the defaults and set cc by the pail just to the left and top of the entrance. colorize was set as follows:
This is the result:
1 Like
When using colorize I set the source mix to zero…this blends the pure color that I choose…then I use the sliders…essentially hsl to pick the color I want……
In this case you have picked
Blended in hue mode but with a 50 50 mix of the input image…which of course you can experiment with but I am not sure really how to control that result to any extent so I set it to zero and blend the color I select controlled by opacity but the mix can be an additional attenuation…
That is my logic…You can use the same logic for some color casts by using the same technique and adding the inverse color of the average of the image although you can achieve much the same now in color calibration or with the color balance module
It just occurred to me they are screen dumps…thats why they dont work…
Your earlier response helped me figure it out. Since I didn’t know where exactly you used the cc color picker, I had to play with the temp and rgb settings to match your jpg. I matched a spot between the S and P in Sprouts using color picker. Of course, it would have been faster if I had realized sooner I needed to switch to softproof to get the color picker to be using the right profile. By and by, I got it. So, thank you very much! 
I got the idea from a gimp snippet.
They get a smoky dusty pastel look using two layers
One a light pink blended in hue at 30% and a second layer in a slightly darker pink blended in difference mode 8-12%…It gives a sort of film look to the image…I think you can get some thing similar by shrinking the dynamic range to give the washed out look and then play with color…
I don’t know if I have a buggy version, but if I moved lightness from 100% to 99.90%, the color changed dramatically. To match your result, I had to have it at 100. Moreover, changes to source mix and saturation had no effect whatsoever.
Strange…blend in normal first and you will see the color you are applying and the effect of the selections have…you won’t fully appreciate it in hue blend mode…at 100 % lightness you are not using the same color I did for sure??
I was just quickly playing around and as I said grabbed the pink from the Gimp article…
As for souce mix…maybe it doesn’t have the same impact as you are blending in hue but still at 50% ie the default you are blending this which is not uniform
and at zero you blend a solid color…
With lightness at 100%, only opacity has any effect. Changing hue should make obvious changes, but it has no effect. Drop lightness down to 99.90%, and lower, and hue has some effect, but it is not what you would expect. Very erratic changes. As I say, something must be buggy. I’ll compile a new version this weekend.
Edit: after compiling latest version/same results:
Look at the color picker rgb and lightness.
Ya I don’t know much a about hue blend mode…could be a nuance there. I would suspect changing lightness maybe should have too much affect. Basically in this module if I use it I usually use it in overlay or softlight as a color filter so I am setting hue sat and lightness to set the color…blend at zero source to blend a solid color…most often I have used it with opposing colors to work on a look or color cast…so there I just tweak opacity to dial in how much of the solid color I am blending…I had actually never used it in hue mode…How to Create a Natural Pastel Photo Effect in Photoshop - FilterGrade.pdf (2.5 MB) I got the idea here…
@priort Thanks for the pdf. I looked at it, and it has potential. In fact, I didn’t know colorize existed, so I have a new toy to explore.
This is what I used to use it for…this or you can try the same also with split toning as you can select colors there…in split toning to do this I would select both colors to be the same…
improves the masking GUI to make the guts of the module more transparent. See here the decomposition shadows vs. highlights:
You also get the graph of opacity vs. luminance for the 3 masks :
At that point, I thinks it’s feature-complete, next step is to write the doc.
It’s been tested and corrected extensively during the past 4 months, maths fixed, and real-world test showed that it’s really fast to get good results (20 pictures edited in an evening, mostly copy-pasting) :
19 Likes
Excellent! I have used this module on a couple of series of pictures, loving it so far!
I am using darktable-3.5.0+1600~g6727a3ab9-win64.exe, but it doesn’t seem to have the new version of color balance rgb. How did you get it?
Really nice work! I’m currently wondering about the new order of
- Global offset
- Shadows lift
- Highlights gain
- power (midtones)
on the 4 Ways tab since the rework from yesterday.
In the master tab it is “Global - Shadows - Midtones - Highlights”. Masks tab is also Shadows → Midtones → Highlights.
Is there a reason for the reordering? It’s no big deal but I’m guessing this might confuse people.
Yes. They appear in GUI in the same order they are applied on the pixels. Highlights and shadows lift/gain are applied on a masked image, so they are truly highlights and shadows. The power, however, is not masked and while it has more visual impact on midtones, it’s applied on the whole range. With this in mind, I find it easier to set them in their order of application on the pixels to avoid circular corrections.
2 Likes
Okay, this sounds reasonable. Thank you for the clarification!
@anon41087856 : Nice work and the only doubt I currently have is already asked and answered (shadows/highlight/midtones order).
Had a quick go at this one a it seems to be much smoother and thus easier to use and correct than the current colour balance module.
@Underexposed:
You can check yourself when a Pull Request (new/changed functionality) has been merged (made available) into the development branch: Go to the darktable section on GitHub, click on Pull requests, click on Closed and set the sort order to Recently updated
In this specific case you see, near the top, Colorbalancergb: rework the mask method. If you hover over the:
#8620 by aurelienpierre was closed 19 hours ago
part you see the exact date/time it was merged. In this case Apr. 8, 2021, 3:59 PM UTC
I’m assuming you build your own (based on this): If you build your version before the afore mentioned date/time you need to pull the changes, rebuild and reinstall.
At the moment I’m on 3.5.0+1616~ge9302dbf7 and it has all the latest changes mentioned above.
How do you learn this stuff??!! Are you a programmer? Anyway, thanks for the info.
Yes, I built mine about 7 hours before it was merged. Time to rebuild.