Just a quick note.

Over at Instagram, Floris posted a recent plot that caught my eye: a Rorschach of red blobs and blue blobs (filled with Hilbert Curves) that overlapped in various places forming purple blobs:

• https://www.instagram.com/p/C61IUtRMri5/

It wasn't clear to me whether he had just used the two colors of red and blue, with the purple forming when the pen's ink blended on the page, or whether he had used three colors to create the illusion of a blend. So I asked and he was nice enough to respond quickly, saying this was an actual blend of only two colors.

This got me wondering how well my blend node could predict that particular shade of purple. So I took a screenshot of his painting and sampled the approximate hex codes of all three colors. The colors in the screenshot are approximate because the overall impression of color is formed by many different pixels of slightly different shades.

I then fed those colors into a simple Venn diagram, using my blend node to fill the intersection, and added a square in the middle with the sampled purple from the screenshot. You can see the results in the attached screenshot and code sample.

You can change the blend mode to see how well the node does. The screenshot shows "Subtract Paint" mode, which comes pretty close. "Subtract Light" mode produces a slightly darker shade of purple which in this case is also pretty close. All these colors are opaque; if you used translucent colors instead for the red and blue you might have some luck with "Translucent" mode, though this would require some work to find the right hue and alpha level, and would also require you to get the canvas color as well. Opaque is easier to work with.

Floris, I wonder if the blend_color node would be useful in helping you predict how different pen colors might blend without having to make a plot to find out for each possible combination. I'd be curious to know how well "Subtract Paint" node would work for this, but there is really no way of knowing without you actually generating many different plots with many different inks and photographing and sampling each one for before/after comparisons. If you are ever bored enough to perform such an experiment, please share the results!

It strikes me that the precise nature of a plotter and the controlled colors of various commercial pens create the possibility of a highly controlled experiments, experiments that would be hard to do with hand-mixed inks and human brushwork.

It also might be interesting to create plots which create the illusion of overlaps, but which in fact substitute a third color for each apparent overlap, with no pens actually overlapping at all. This could be done with an ordinary compound node, as I did here, or, in the case of multiple overlaps, by using my overlaps and recolor nodes as shown in the library demo for my blend_color node.

That's it. Comments welcome. My complements, again, Floris, on your lovely plot.

John