On light, colors, mixing paints, and numerical optimization
github.com
github.com
I've been getting into mixing for miniature painting myself, but with a bit of added perspective which (I think) isn't captured here.
It turns out there's more than just raw color going on here. Acrylic paints have different degrees of opacity, and you can increase transparency by mixing in acrylic medium. Additionally, adding water and/or flow-improver helps the acrylics slide off flat surfaces into crevices, producing deeper colors in recesses and thinner applications in open areas - this is the principle which gives us washes, shades, and contrast/speed paints.
And there's basic acrylic mediums (matte or gloss) but also specialty mediums which have more texture (gels, modeling pastes) or additives (pumice, fiber, glass micro-beads). [Edit: And of course, Metallics!!! There's a few different additives, like mica, which are used to create metallic paints.]
Together, this points to a combinatorial problem: No pre-mixed bottles will ever cover all of the possibilities, so you have to start mixing paints (and mediums!) to achieve the desired effects in any particular project.
Also Bruce MacEvoy's in-progress book "Color Experience"[1] looks like it will be a fairly definitive treatment of the topic. It is also rooted in physical colorants (the author is an extremely serious watercolorist).
Sometimes I think combinations will have _more_ of an effect - I was hopeful that ink on paper below gloss medium acrylic would be more obvious, but did not come out the way I expected.
He complains a lot about lightning in museums, that it deviates a lot from daylight, while many works of art are create by daylight. The light is often too yellow, based on some 'dubious' research that museum visitors prefer yellow light. LED lightning often causes mixtures of pigments that look the same under daylight to look different under LED lightning. LED lightning often is rather weak in the infra-red to red part of the spectrum cause most reds in painting to fade. He found a combination of a reddish and a blue TL-lamps supporting to LED lights to give a better approximation of daylight. When this was applied to some museum rooms, the son of a man who donated a work, asked what they did with it work as it looked again like how it looked at their home when it was hung close to the windows. (Shortly after this the museum director decided to use only LED lightning for the new wing that was going to be build.)
The problem of course is that we are so used to adapt our color vision to the lightning that most people do not see the difference. I guess that people who mix colors all the time, develop some kind of absolute color vision and do notice the effect of improper lightning on art works.
[1] https://www.pstruycken.nl/EnDyn.html?Li,tag=q&w
I'm surprised that this extremely cool phenomenon is not embraced by artists and exploited as an essential part of their language. For example, they could use pigments with bizarre absorption spectra/fluorescent properties, so that the same painting looks completely different when put near a sunny window or under a led lamp.
Focusing only on color rendering under daylight seems a bit short-sighted, now that we master all sorts of synthetic lighting and fancy chemistry.
Imagine a Mona Lisa that only smiles in a cloudy evening!
But what about the opposite? A sunny landscape when it's cloudy outside, and a cloudy landscape when it's sunny?
The spectrum data gets boiled down (or not) in a couple of ways:
1) Rf, or “fidelity index” where a score closer to 100 means it renders colors more similarly to how they look under a reference illuminant (which standard illuminant is used for comparison depends on color temperature). But as you summarize more information down to fewer values, you of course get less detail - many different spectrums could provide an Rf of 95, with different visual effect. Rf only gives you an idea of whether the visual difference between this light source and the reference illuminant will be large or small. It’s a similar concept to “CRI” scores if you’re familiar with that, but based on a much better set of samples, and not as easily gamed for a high score by carefully designing an LED’s spectrum to correctly render CRI’s more limited set of color samples.
2) Color vector graphic, shown on page 28 of the PDF linked below, shows a smushed circle which broadly illustrates how different hues are shifted. For example, reds might be undersaturated and shifted toward orange. This is still summarizing, not all red objects will have the same shift, depending on their precise spectral reflectance distribution and the source’s spectral power distribution. But it gives a better idea of how the light will look than the Rf alone.
3) Color sample vectors shown on page 24 of the same PDF, which shows each of the individual 99 spectral reflectance samples used for TM-30 and how it will shift compared to when viewed under the standard illuminant. The color vector diagram in point 2 is a summary of these, but if you want more detail you could see them individually. I’m not sure anyone actually publishes this since these particular 99 samples aren’t likely applicable to whatever you’re doing. And ultimately it’s still not the full story, because this is all based on 99 specific color samples attempting to be a representative spread of materials you might encounter in the real world. But 99 samples certainly can’t represent all materials. To get real answers you’d need to go down the same rabbit hole that the article did and measure the spectral reflectance of every object you’re lighting.
https://www.energy.gov/sites/default/files/2022-04/ssl-royer...
Compared to incandescent lamps, fluorescent sources including LEDs emit spectrum that is bumpy or even spiky. Those spikes fall within R, G and B and the resulting light looks white itself; but the objects or paints from which it reflects may not look the same[0]. At extremes, depending on what the paint is made of (not simply its colour!), the picture may look very different.
(Ever noticed how you may admire an amazingly vibrant flower in the field, but no photo seems to reproduce that brilliance? That’s another example of metameric failure.)
To see a work of art how it was intended to look, you should view it under the light it was created in. Unfortunately, it appears impossible to break from the shackles of LED these days. Depending on where you live, it may actually be impossible to source a tungsten filament bulb. You’d think it should be easy and worth the extra electricity expense to the management of any good gallery or museum, but apparently not.
[0] See: https://en.wikipedia.org/wiki/Metamerism_(color)#Metameric_f...
There are simple ways to conceptualize color mixing, for example the primary model, or HSV. Of course, those are models, so they don't capture everything, as mentioned ITT (e.g. transparency, lightfastness, texture, potency, price) but they're practically good enough.
For example, in a primary (+white) setting, a given mix can only:
- not have enough red
- not have enough blue
- not have enough yellow
- not have enough white
So the only questions one can ask is: "does it need more red/yellow/blue/white?", period. If one has more than those 4 colors, then the trick is to understand them in terms of red/blue/yellow/white: one is then left again with those 4 questions.
I've found HSV more practical than RYB, but people are generally accustomed to RYB, hence why I've described it instead; the general idea really is the same: only a small, finite number of mistakes.
It's simple, but it still requires a few hours of practice until it becomes intuitive, especially with "real world" pigments.
Here's a short video of an older version. Newer one let's you set how long to search for.
Light is therefore color, and shadow the privation of it by the removal of these rays of color, or subduction of power;
and these are to be found throughout nature in the ruling principles of diurnal variations.
The grey dawn, the yellow morning and red departing ray, in every changing combinations,
are constantly found to be by subduction or inversion of the the rays of their tangents.
-- J. M. W. Turner, 1818.
J. W. Goethe, preface to The Theory of Colour, 1810