Aurelien Pierre #aurelienpierre could you watch the last 18 minutes of this video and help translate the techniques that are being shown into darktable modules? I have been trying with color balance RGB but I feel that I am missing something.
So, if you want to isolate the hues in an image like she did in the first part of that video, what you can do is disable filmic, enable base curve with preserve colours=luminance, and set the curve horizontally flat at around 50%, which will remove contrast and just leave the colours. You can then use global colour picker in LCh mode to identify what are the hues in the image.
In terms of pushing hues around (eg. making blues more cyan), you can do this with color zones with the “adjust hue, select by hue” mode. If you want to affect the tint in highlights and shadows, then this can be done on color balance rgb 4-ways tab (eg. put more blues in your shadows and more yellow in your midtones/highlights). Note that the latest darktable has a vectorscope which may be useful (click on the mode button in the darkroom histogram to cycle through to it).
Note that color zones works in display-referred space. If you want a scene-referred alternative, the new colour balance RGB module can do it. There is a hue rotation slider. You would probably want to combine it with a parameteric mask with feathering to limit the scope of what is is affecting in the image.
Great suggestions, thanks!
For a basic/simple test I generated a palette with Paletton based on color grabbed with your suggestion (base curve). ​
Then I used color balance rgb: in 4way tab I selected the hue from Paletton and with a sequence of try with chroma and luminance sliders I emulated the palette (roughly).
Technically it looks ok… now, the hardest and longest part, educate my taste for colors and the philosophy of color grading…
1 - Chose a base color. Since the main subject is my son, I chose his face’s hue. I followed @Matt_Maguire 's tip (thanks for that) and picked the hex value on my son’s face:
3 - Then it was the time to define which areas in the image would be influenced by which palette color. I chose his shirt to fit the bottom left hue, the wall, the bottom right one, and the pillows (?) the top left. I got the hue values following @gi.vi.23 's tip, above.
Then I created masked instances of color rgb to affect each area of the image, as seen in the after screenshot above, set the hue values and tweaked luminance and chroma, but also tweaked perceptual saturation in the instance affecting the pillows, to reduce their dominance.
And that’s it.
Not a big change, but I think I could make my son’s face pop out a bit, by adding some color contrast/harmony to the image.
Not that I necessarily prefer the graded version, but it was certainly fun to play with that.
The Tetrad works as 2 pairs of opposite colors. In your picture would be the red opposite to the green and the orange to the blue.
But in your picture, the orange (skin) is in the same area as the red (pillow). That’s why I understand that doesn’t contrast.
If the model would be standing up to the wall would be different.
Anyway the workflow is right. Thanks for sharing!
You can use the vectorscope in the dev version after doing the trick of the basecurve. So, you can see where are your colors too