PDA

View Full Version : Letting the computer calculate paint recipes


Christian Skeel
03-26-2014, 07:29 AM
If you would like to try getting recipes for mixing paint by just pointing on the color on an image in your computer you can get this service online for free here:

http://sensuallogic.com/paintmaker/Online.html

The system works with: Rembrandt Fine Artist's Oils, Golden and Liquitex acrylics.


http://www.wetcanvas.com/Community/images/26-Mar-2014/1730221-ScreenShot.jpeg

http://www.wetcanvas.com/Community/images/26-Mar-2014/1730221-ScreenShot2.jpeg

Cheers Christian

Mythrill
03-26-2014, 11:43 AM
Hi, Christian. Recently, I have been studying this painting from Ambrosius The Elder to paint:

http://www.wetcanvas.com/Community/images/26-Mar-2014/96427-Bosschaert_the_Elder,_Ambrosius_-_Still_Life_with_Flowers_in_a_Wan-Li_vase_1619.jpg

This is from the Dutch Age golden era of painting, and by studying it, I suspect the artist used vermillion, madder (maybe even two qualities, "red" madder (natural Alizarin) and "rose" madder.) If you use this particular photo and color-correct it to "remove" the varnish, you will also see that the artist used a lot of blue, too. The curtains are actually all blue possibly even Ultramarine, which was expensive back then!

In my case, I selected Golden's Acrylics. All the paints, cold daylight. This is what I get:

http://www.wetcanvas.com/Community/images/26-Mar-2014/96427-Recipe.jpg

Recipe 1 obviously doesn't work, and neither does recipe 2. Recipe 3 would be the closest, but I raise an eyebrow when I see "red oxide" listed there. I painted the pink rose in the vase with red oxide and white, which is indeed close, but PV19 + PR 233 + PR101 (transparent) and a filler, to simulate the weaker madder lake, plus PBr7 for the shading, was what gave me the closest results.

Also, there's a lot of yellow in that painting. Where's the yellow in the 3 recipes? And I don't think Dioxazine Purple (PV23) works at all to reproduce this painting!

In a nutshell, I consider this program very insightful to try to guess colors, but nothing replaces actually trying to guess the hues on your own by studying the painting with small miniatures (and judge what's best for you.)

Christian Skeel
03-26-2014, 01:22 PM
The recipes will give the rgb values requested and you can select or deselect the pigments you want to work with in the set and in that way force the program to find recipes with the pigments you like. I made the program because I live in Denmark where we do not have enough daylight specially in the winter month. Without the proper light I made too many mistakes because I wasn't able to jugde properly. I also wanted to to try working with all the paint for a painting ready and wet at the same time in order be as free as possible with the brush strokes. I use PaintMaker for many of my paintings and enjoy being able to work fast.

Patrick1
03-26-2014, 07:36 PM
This is interesting Christian, and thanks for sharing but I have to ask: does it/ how does it deal with the varying color of transparent paints from masstone to undertone?

For example Prussian Blue or Dioxazine Purple are virtually black in masstone or Nickel Azo Yellow PY150 which is brown in masstone but bright yellow in undertone, Quinacridones are somewhat maroon in masstone. Transparent pigments don't have just one intrinsic color, it's a whole range depending on how it's applied. Thanks.

Christian Skeel
03-27-2014, 12:28 AM
Hi Patrick
The system only works with masstone. You are right undertone cannot be calculatet because there are no way of knowing the precise result. In Paintmaker there are always one masstone color present in the recipes. Usually titanium white. So the result of the recipes always comes out as masstone. I find it practical to work with since I want to be in control when working with transparant colors.

waterhousey
03-27-2014, 12:53 AM
While entertaining, I find these virtual paint mixers to be practically useless. Golden's "Mixr," specifically, would often tell me I ought to use colors that I don't feel are necessary in my paintbox, i.e., Light Ultramarine.

Really, Golden? Light Ultramarine? I love your paint, but COME ON.

joe publik
03-27-2014, 08:47 AM
Hi waterhousey

"While entertaining, I find these virtual paint mixers to be practically useless".

Not so in this case. The major paint and auto companies have been doing this for ages. Golden's mixing program, like many that have appeared over the years, is not very good. Christian's model derives from Zsolt Kovacs drop2color and color2drop. Christian freely acknowledges this. Zsolt's models are rigorous and they work, subject to certain boundary conditions such as masstone and transparency. Accuracy can be spectacular and will certainly be as good as an eyeball match over most of the gamut defined by the range of paints chosen.

Whether these techniques are useful or of interest to all painters is entirely another question.

Mythrill
03-27-2014, 10:54 AM
Hi waterhousey

"While entertaining, I find these virtual paint mixers to be practically useless".

Not so in this case. The major paint and auto companies have been doing this for ages. Golden's mixing program, like many that have appeared over the years, is not very good. Christian's model derives from Zsolt Kovacs drop2color and color2drop. Christian freely acknowledges this. Zsolt's models are rigorous and they work, subject to certain boundary conditions such as masstone and transparency. Accuracy can be spectacular and will certainly be as good as an eyeball match over most of the gamut defined by the range of paints chosen.

Whether these techniques are useful or of interest to all painters is entirely another question.

Hi, Joe!

I've read Christian's new post. It seems it was him who coded the program, so I apologize if I was harsh. While I still think the program will not give you a perfect recipe for a paint (certainly no better than studying it yourself and doing miniatures to check what colors match best what you want,) coding something like this is very difficult, and his efforts are amazing.

As for car industries, it works well for them because they use a paint in full masstone. It's not a nuanced painting like a handmade one of a landscape and / or a portrait is, where the artist can easily use 12 pigments in one single paint (if he has them) and and tint and mix between those colors inbetween!

If you only want the masstone of a paint, one cheap alternative would be to compare a certain spot of the paint with the masstone of a certain paint. For instance, if you suspect that Cerulean Blue (PB35) was used in the sky, pick a point in it and use an image program (like paint) and it will give your RGB values. Then, get a photo of a Cerulean Blue and pick a point in it as well. If the values are close, that pigment was probably used.

Of course, this method is not accurate if you don't know if Cerulean Blue was mixed with something else, like any yellow, white, black, or something else (or if there's a glaze, etc.) But you could still attempt to use any other set of colors to reach the RGB values of that point in particular.

joe publik
03-27-2014, 11:34 AM
Hi Mythrill,

"...compare a certain spot of the paint with the masstone of a certain paint..."

I think I know what you mean. I ssume you are suggesting an interpolation from close colors (if not I apologise).

It's not as simple as that because you are assuming that linear interpolation will work. It won't and that's why programs like Golden's fail. Even Photoshop fails. Handprint does a good job of showing you how color mixes mostly run along curved paths (in 2 dimensions). David Briggs is also good. But 2 dimensions are not enough. You need 3. Christian chooses R, G and B. I prefer L, a and b. The genius of the original programs by Zsolt (years in the preparation) is their ability to perform this interpolation and prediction in 3 dimensions.

Some people have an eye for color just like some people have an ear for music. I don't and it is wonderful for me to be able to mix a color, any color, in 2 minutes. And the colors are mostly correct or "good enough" - I check them with an XRite spectrometer. If they need a final touch with red, or green or blue then I can do that too in a highly predictive way.

But, as I say, it's not for everyone.

Christian Skeel
03-28-2014, 12:03 AM
The mathmatics behind Zsolt Kovacs program Color2Drop which I use as the calculator for my program is made by Kubelka and Munk. It is the only way calculate colour mixes from lab og rgb coordinates to many pigments. I have tested the program for years and found no wrong colors. It is amazingly precise. But there are things you have to know to avoid getting off colours. I explain how to prepare your image before measuring and mixing in menu 'tips' in PaintMaker. When the image is within the colorspace of the pigments used the recipes is almost boring acurate. It takes me 30 to 45 minutes to mix 30 colours precisly. Nothing compared to the time I was used to. Now I can paint with all the paint wet and ready mixed and therefor make brush strokes in a totaly different maner.

Christian Skeel
03-28-2014, 12:23 AM
Here are a test painting for a portrait compared to the photo I took:
http://www.wetcanvas.com/Community/images/27-Mar-2014/1730221-fotoIMG_3173v2.jpeg
The photo of the painting is taken with my iPhone so the colours differ from the photo but one can see the color harmony is the same between photo and painting. And you must take my word for in real life the colours match perfect. This painting took me 3 hours to make including everything. So now I often make test paintings to see if I am on the right track. In this case the method wasn't good and I have decided to paint the portrait in another way.
http://www.wetcanvas.com/Community/images/27-Mar-2014/1730221-maleriIMG_3803.jpeg
And three ways to paint the same image made possible because of the speed and precision in the preparations:
http://www.wetcanvas.com/Community/images/27-Mar-2014/1730221-1Remote00099-EditFIN.jpeg
http://www.wetcanvas.com/Community/images/27-Mar-2014/1730221-Remote00079-EditFIN.jpeg
http://www.wetcanvas.com/Community/images/27-Mar-2014/1730221-Remote00083-EditFIN.jpeg
And another way to use the brushes made possible by this technique:
http://www.wetcanvas.com/Community/images/27-Mar-2014/1730221-IMG_3636.jpeg
http://www.wetcanvas.com/Community/images/27-Mar-2014/1730221-IMG_3619.jpeg

jorri
05-02-2014, 09:30 PM
I'm interested in this, i wouldn't use it all the time necessarily, (as sometimes i don't paint exact photo colours) but i've done experiments about this mostly from handprint's nformation and it really helped my paintings.

I understand its more for your personal use but I wonder if there's a way to add more paints though? from the rembrandt set only titanium white, ivory black and yellow ochre i have with the white being used most frequent, and plenty which aren't very common at all.

Also only a couple of suggestions, i have no idea how i would program somethng like this and by the looks of it it works very well, but:

-are more than 4 pigments necessary? Given colours are three dimensional you should only need four for any colour, making a tetrahedron.
-perhaps this is due to the gamut of paints, I think maybe adding a 'widest gamut' style pallete could make the mixes more efficient.

jorri
05-02-2014, 09:34 PM
I'm interested in this, i wouldn't use it all the time necessarily, (as sometimes i don't paint exact photo colours) but i've done experiments about this mostly from handprint's nformation and it really helped my paintings.

I understand its more for your personal use but I wonder if there's a way to add more paints though? from the rembrandt set only titanium white, ivory black and yellow ochre i have with the white being used most frequent, and plenty which aren't very common at all.

Also only a couple of suggestions, i have no idea how i would program somethng like this and by the looks of it it works very well, but:

-are more than 4 pigments necessary? Given colours are three dimensional you should only need four for any colour, making a tetrahedron.
-perhaps this is due to the gamut of paints, I think maybe adding a 'widest gamut' style pallete could make the mixes more efficient.

p.s. the ambrosius painting posted is completely the wrong method, you have analysed one colour selected, and it looks pretty accruate to how i would mix that colour and i frequently mix it. It doesn't analyse and separate the whole painting, that is why you don't have yellows and blues-because you have to click on them.

Mythrill
05-03-2014, 10:03 AM
-are more than 4 pigments necessary? Given colours are three dimensional you should only need four for any colour, making a tetrahedron.
-perhaps this is due to the gamut of paints, I think maybe adding a 'widest gamut' style pallete could make the mixes more efficient.

Hi, Jorri.

To reproduce a wide gamut of colors, only 3 paints are necessary. That's proven and derived from the paint industry. Black is only added for more chromatic purity (i.e, "deeper blacks") and saving paint (you print a lot of black text in your computer, and the other 3 pigments are always more expensive than black.)

Printers usually use the pigments in their dye form, and here they are (codes listed:)
Yemon yellow: PY 1, PY 3, (lemon shades) PY 73 or PY 74 (medium shades.) PY 1 and the lower grades of PY 3 (worse lightfastness) are used for cheaper printer paint.
Cyan: PB 15:1 (red shade,) PB 15:3 (blue shade.)
Magenta: Lithol Rubine (PR 57:1,) Quinacridone Magenta (PR122, higher-performance printings.)
Black: Possibly Carbon Black (PBk 9.) This one is a pigment, though, so either some sort of special preparation is made, or your printer will use Aniline Black (PBk 1,) which is a true dye.The problem is that printers are much better than us regarding mixing colors, and can juxtapose then in dots with microscopic precision, invisible to the human eye. Even then, they can't mix the gamut of colors you can in real life. This can be seen quite clearly on the whole violet range, violet-blues, turquoise greens, and yellows (what range, exactly, depends on your printer.)



p.s. the ambrosius painting posted is completely the wrong method, you have analysed one colour selected, and it looks pretty accruate to how i would mix that colour and i frequently mix it. It doesn't analyse and separate the whole painting, that is why you don't have yellows and blues-because you have to click on them.
That method doesn't analyze the whole painting because, in real life, artists particularly on the realist side tend to mix several different colors and put them together to achieve a nice painting. You can reproduce the base, for instance, by mixing yellow ochre, ultramarine blue, black and white, but in all different proportions depending on the spot you're working on.

That's why reproducing a painting is so difficult in the first place. There's no exact way to know what pigments, in a certain proportion, were mixed to create that look in one particular point of the painting. There's more to consider, indeed:

Did the artist use mixed media? Different pigments look different on different mediums, and the artist may have used different mediums to achieve the result (s)he wants. One example in classical painters is the usage of egg tempera as an underpainting.
Did the artist use glazes? Glazes can be simulated on your computer screen, but they can't be perfectly reproduced. The reason is the phenomenon of paint glazing is that light bounces through two or more physical layers, trapping light and sending them back to your eyes. Monitors are just one surface, and they emit light, instead of reflecting it.
How was the pigment the artist used manufactured? Even pigments like the obscure genuine "Naples Yellow" (PY 41) can be reproduced in ranges from "lemon yellow" to "orange-yellow," which will give them unique handling properties and characteristics. This is also why some artists consider their "ideal" Alizarin Crimson (PR 83) either a deep, bright crimson red, a reddish rose not too different from rose madder, or a dull red: that's because there are around 18 variations of it, only around 3-4 being relevant to painters.If you use color pickers around certain objects, you'll usually get neutralized versions of the actual pigment the artist used. To get a closer approximation, you would have to use your color picker at least on the lightest spot of an object, the "medium" spot, and the "dark" spot. Even then, due to the complications listed above and a few more you still can't exactly know what the artist used without deeper analysis (x-ray and microscopic analysis, for instance,) and/or historical references.

jorri
05-03-2014, 12:58 PM
i mean 4 colours in the mix, not a whole palette of four colours.
printing uses 3, but can get more with 6 etc. but that's just the pallette.
It also uses blank paper, which painters don't-they use white, so in reality its 5+ pigments, often up to 8 or even more for professional printers.

By a wider gamut, i mean higher chroma paints, with higher chromas of different values and earth colours in addition to it. There is value, chroma and hue so this makes a 3d 'colourspace'. You can't make a 3d object with 3 points so to get colours in between you'd need a fourth point, and shouldn't need more because that shape would cover the whole gamut- painters don't paint like this but its an opportunity if there's a program to calculate it for you.

an example is you can mix an inbetween hue from two high chroma paints, then desaturate at an effective value with white and black.
Or perhaps balance a higher chroma mix in the value dimension by using darker/lighter high chroma versions of the colour to adjust value.

--
the ambrosia painting i meant- you posted one colour from the painting and compained there's no yellow or blue- you don't put the whole painting through the processor to get various colours- you select individual colours to pick different parts of the painting...you must have selected just a red, which it seemed to mix exactly how i would of the 'picked colour'-not necessarily how the painter did it.. But i recommend selecting different parts of the painting because the flaw was in how you selected the colours.

I think its more meant for a photograph rather than analysing the make-up of a pre-made painting, i mean you could accurately produce the photo of the painting-like art prints are made- but not the painting itself because that is more complex.

jorri
05-03-2014, 01:01 PM
has anyone managed to get color2drop by the way?
there is no download link, just an email so not sure who its sent out to...

Mythrill
05-03-2014, 02:48 PM
i mean 4 colours in the mix, not a whole palette of four colours.
printing uses 3, but can get more with 6 etc. but that's just the pallette.
It also uses blank paper, which painters don't-they use white, so in reality its 5+ pigments, often up to 8 or even more for professional printers.


Hi, Jorry!

Yes, printers only need 3 colors; however, in practice, they use 4 to create a few interesting mixes the CMYK system.

One example of black mixing in printers is mixing black to cyan. The black pigment in printers is naturally a bit dull, and this is doesn't look attractive in commercial prints.

There are several "recipes" for your printer mixing "rich blacks" even those on the warmer side of the wheel (i.e, not needing cyan at all.) You can see some at the Wiki:

http://en.wikipedia.org/wiki/Rich_black

Another interesting usage for black in printing is to simulate the hue of gold. When mixed with black, yellow shifts to green, and when the dots are mixed with yellow + magenta (in different proportions, to create yellow-oranges,) you'll get a your gold color.


You can't make a 3d object with 3 points so to get colours in between you'd need a fourth point, and shouldn't need more because that shape would cover the whole gamut- painters don't paint like this but its an opportunity if there's a program to calculate it for you.


The problem is that painters aren't always worried about exact color matches that's why a few frown on color mixing software. They're more interested in reinterpreting the gamut they see.

Suppose someone is painting a river on an overcast day. Most likely, the sky will be gray from the clouds, and the painter thinks to himself: "you know, what if there was a sunset here instead? What if the sky had some yellows with some rust-like tones?" Said painter then will simply ignore the actual colors of the sky and paint according to his/her personal preferences.

Even classical painters reinterpretate scenes. The painting you mentioned, for instance, has a carnation as the very first object in the foreground. I did some research, and much to my surprise as the painting looks realistic there are no carnations with that exact petal patterns, nor carnations with those exact shapes. Ambrosius did a great job into fooling me that flower actually existed as it is depicted!


an example is you can mix an inbetween hue from two high chroma paints, then desaturate at an effective value with white and black.
Or perhaps balance a higher chroma mix in the value dimension by using darker/lighter high chroma versions of the colour to adjust value.


the ambrosia painting i meant- you posted one colour from the painting and compained there's no yellow or blue- you don't put the whole painting through the processor to get various colours- you select individual colours to pick different parts of the painting...you must have selected just a red, which it seemed to mix exactly how i would of the 'picked colour'-not necessarily how the painter did it..


Agreed. That's probably not how the painter did it, because this particular painter used copper surfaces, for instance, which gives paintings a natural "enamel" finish. He also made his own paint, which had completely different characteristics.

What I'm trying to do is to try to find single-pigments (or hues) that would best imitate his pigments sometimes dulling them down and adding fillers, if necessary, so the pigments will look a bit less saturated and more transparent.


But i recommend selecting different parts of the painting because the flaw was in how you selected the colours.


Thanks. I might try these alternate methods to improve on color detection. I wouldn't completely rely on them, though, exactly for the reasons you mentioned below.


I think its more meant for a photograph rather than analysing the make-up of a pre-made painting, i mean you could accurately produce the photo of the painting-like art prints are made- but not the painting itself because that is more complex.

Indeed that's one problem with photographs: since they usually are photos and prints from a limited color gamut and a gamut almost never completely reflects the colors from original paintings you're restricted to duplicate the photographs with 100% accuracy on canvas, at most.

Since I don't have this painting, what I'm trying to do is to look at several different photos that will reflect different hues of the painting. Different lighting and cameras will reveal some hidden hues. Here, at home, I'm using a darker photo, and it shows no blue at all.

The photo I sent to this forum, on the other hand, shows some blue on that painting. After color-correcting this particular photo on Photoshop (to have an idea of how it would look with a clear varnish) I discovered the painting uses loads of ultramarine blue (and/or another blue pigment, like azurite.) There actually are curtains on the background that are completely blue!

So, yes. Looking at different photos of a painting is a less-than-ideal solution, but it's the best resort you have if you want to try to duplicate it and have an idea of how it looked in the past plus, you learn a lot too. :)

joe publik
05-05-2014, 08:11 AM
I think (but of course am not sure) that the very best available copies of originals are those available on the Google Art Project site. They will have doubtless spent a fortune on their technology so I'm going to assume that their color rendition will be the best available. Certainly their resolution is more than enough to allow (in most cases) more than adequate resolution for a home copy. Reproduction is allowed for personal use.

Christian Skeel
06-09-2014, 07:58 AM
The link to PaintMaker color mixer has changed:
http://sensuallogic.com/paintmaker/OnlinePaintMixer.html