All my life I've seen pro painters talk about "values" by desaturating in Photoshop, and turns out it's one of the most widespread mistake: this is the wrong way to assume values, as the maths are wrong, and let me show you why:
Desaturating in Photoshop doesn't accurately portray brightness at all.
A perceptual model is much more accurate to the way we perceive brightness & contrast.
It's seems subtle, but when toggling between the two conversions, it's clearly not something to be overlooked. Photoshop's desaturation results in a incomprehensible mud of grays which has nothing to do with the reality of color science.
Let's desaturate primary colors:
- Red
- Green
- Blue
And secondary colors:
- Magenta (R+B)
- Yellow (R+G)
- Cyan (G+B)
The result? A uniform 50% gray. Non-sense.
- Red
- Green
- Blue
And secondary colors:
- Magenta (R+B)
- Yellow (R+G)
- Cyan (G+B)
The result? A uniform 50% gray. Non-sense.
Why desaturating pure colors ends up with the same gray in Photoshop? On screens, we're obviously in RGB with additive blending, so secondary colors (Cyan/Yellow/Magenta) should be brighter than primary ones, as they reflect more incoming light. Check out this GIF I just cooked:
According to the picture up there:
- pure primary colors should end up 33% gray.
- pure secondary colors should end up 66% gray.
Let's try this naive conversion using the channel mixer, each primary color contributing 33% of the final output.
- pure primary colors should end up 33% gray.
- pure secondary colors should end up 66% gray.
Let's try this naive conversion using the channel mixer, each primary color contributing 33% of the final output.
This results in something correct mathematically as expected, but it still looks flawed. We still can't discern hues at all, while we should be able to.
Why? Because this was a naive conversion.
We forgot that our human visual perception isn't equally sensitive to red, green & blue. We are way more sensitive to green than red, and red than blue. The more sensitive we are to a color, the brighter it looks.
We forgot that our human visual perception isn't equally sensitive to red, green & blue. We are way more sensitive to green than red, and red than blue. The more sensitive we are to a color, the brighter it looks.
Our ancestors navigated dangerously through foliage, had to gauge forage & fruit ripeness from afar: those with better vision for these tasks survived more often - natural selection led to our specific blend of trichromacy with 3 types of cone cells.
sciencemag.org/news/2017/02/y…
sciencemag.org/news/2017/02/y…
So what we need is called a "Luminosity function".
It's not perfectly accurate, but it's a valuable approximation of the human perception of brightness across the color spectrum.
wikiwand.com/en/Luminosity_…
&
wikiwand.com/en/Relative_lu…
It's not perfectly accurate, but it's a valuable approximation of the human perception of brightness across the color spectrum.
wikiwand.com/en/Luminosity_…
&
wikiwand.com/en/Relative_lu…
Here's the common approximation:
Luminance = red 21% + green 72% + blue 7%
Yes, we really see green roughly 3x brighter red, and 10x brighter than blue. We can easily model the contribution of each primary color in Photoshop using the channel mixer.
Luminance = red 21% + green 72% + blue 7%
Yes, we really see green roughly 3x brighter red, and 10x brighter than blue. We can easily model the contribution of each primary color in Photoshop using the channel mixer.
Much better.
Now we can discern hues in black & white, and often we can even guess the original color with some experience.
Now we can discern hues in black & white, and often we can even guess the original color with some experience.
Let's try this on the original picture, and see how natural & correct it feels. It's almost like there is no change in brightness.
It's now obvious how the Desaturate method destroys perceived brightness. When we work with values, we need correct insights to make our decisions.
This is why you probably saw a lot of teachers telling you about "values" and it never felt like a reliable technique to you.
This is why you probably saw a lot of teachers telling you about "values" and it never felt like a reliable technique to you.
Now that we understand some basics, bonus rounds
-------------------------------------------------------
We can let Photoshop approximate the work for us:
Trick #1
- Create a black layer on top
- Set its blending mode to "Color"
- That's it. You can toggle it on/off.
-------------------------------------------------------
We can let Photoshop approximate the work for us:
Trick #1
- Create a black layer on top
- Set its blending mode to "Color"
- That's it. You can toggle it on/off.
Trick #2
-------------------------------------------
- Setup a custom color proofing
- Select "Gray Gamma 2.2" on Windows
"Gray Gamma 1.8" on Mac
(sGray & Dot Gain 20% also acceptable)
- Quickly toggle luminance ("values") with CTRL+Y at any time
-------------------------------------------
- Setup a custom color proofing
- Select "Gray Gamma 2.2" on Windows
"Gray Gamma 1.8" on Mac
(sGray & Dot Gain 20% also acceptable)
- Quickly toggle luminance ("values") with CTRL+Y at any time
Such a pleasure to see hundreds of thousands of people mindblown from this thread. Sharing knowledge & unlocking insights is one of the most enjoyable feeling.
A nice reminder of what social media should be - it makes me want to tweet more knowledge, less thoughts & comments.
A nice reminder of what social media should be - it makes me want to tweet more knowledge, less thoughts & comments.
What if I told you that somehow, we're still slightly off the ground truth? 😭 To achieve correctness, we'll have to convert back & forth from gamma to linear- and you'll see all the benefits of painting in Photoshop in such an alien configuration.
Thread update later today 🤞
Thread update later today 🤞
