Hi everyone!
I saw this cool thing that Tom did with that drop-down select!
It really is an awesome development.
To inkman: Yes, it seems as though Tom is attempting to copy
my idea.
But, the crux is that it doesn't matter that the seps are in the browser. Where they are is irrelevant, as long as you can get access to them. Installed software is the fastest as it works with no downloading, which is the slowest system in your computer. Ease of access is something that could be named as a benefit to browser systems... but really, no screen printer is looking to quickly sep when they are at a friend's using their PC. Needing to install is a non-issue.
Moreover, the browser is a specialized piece of software that is meant to present and display, not calculate and run. Now, Google is trying to change that so that more and more things can run in the browser. There are benefits, sure, but Google mainly wants this so they can push their Chromebook/ChromeOS systems. They have ample clout, of course.
As for seps in the browser... I can't do it well. I can do it good enough to produce a smoking preview, that's actually accurate to boot, but the real seps get run on the server in my system. I'm sure Tom took his SimpleSeps and and bolted on a rudimentary web interface. That is an awesome idea. I wonder how he got it.
But here's the thing. It doesn't matter where you get your seps. It matters how good they are. It matter how long it takes you to get them, how easy it is to work that system and of course price. But that it's in a browser, plugin or dedicated app is not an interesting point, at least not until all seps systems become so equally high quality that it becomes the only differentiating factor.
We aren't there at this point. Right now we have a choice between plugins that run PS commands For You and one to three standalone programs that have some actual specialized sep code.
What I did, is I made a software that takes a brand new look at color separation.
Everyone is used to pulling colors. This is a bad paradigm. A color can be pulled 100 times. That's a problem. When I 'pull' a color, the color stays in the image. And I have a 'copy' (inaccurate copy) of it on a sep/film. There's nothing stopping me from re-pulling that color except maybe some common sense. But the common sense doesn't help me when I pull a purple and a red, and the purple pulls some red, and the red pulls some purple. I just puled purple-red twice. By how much? Nobody knows. It's not an accurate answer. It's a guess. It may work well, sure. Maybe good enough even, so we don't have to fuss about it.
But good enough and perfect are quite a separate idea.
Math doesn't make mistakes. So then it becomes the task of the scientist (Isaac Newton, Albert Einstein, Max Planck) to take the physical world, and model it in pure math. Then we can do calculations on the real world. In the screen printing realm, we're talking about inks and emulsion and light, which get modeled as vectors, reflectivity coefficients, partitive mixing math, and other such numbers and models and they become useful if we know how to run calculations on them.
So, there are several ways to get to a result. There's the common sense and trial-and-error approach, which get us amazingly close to our goal, whatever the goal is. It's how we learn how to walk. We try, fall, try some more, and before long we walk. It is also how the noob screen printer becomes a master separator. Trial. More trial. And a whole lot of error along the way.
Then there's the observational method. It's what Isaac Newton used when he devised his gravitational formulas. Pretty observant, this dude. He is one of the greats, for sure. Ultimately, he was proven wrong, however, because his perception was too coarse to find the subtle shifts caused by temporal distortions. And Einstein fixed that for him. Still, even Einstein wasn't quite right. Quantum mechanics has proven some of his ideas to be wrong, although his relativity theories have held up quite well. We can measure time shifts in airplanes with atomic clocks. As unintuitive as it seems that time isn't the time we thought it was, it apparently is true. We've measured it.
Now, the trial and error way will get you close every time, given enough trials and errors.
But math doesn't allow for error (unless it does, in which case it's very precise about the error indeed). The colors we see can and are often modeled in math. Numbers of amplitudes and frequencies. Perceptual models that describe how we see. Models of reflectivity, absorbency, refraction, fluorescence and phosphorescence to describe properties of inks and substrates.
Given the eye's ability to only distinguish between red, green and blue due to the 4 different light-sensitive cells in our retina, it is straight forward to model all colors as RGB. Some kind of RGB. Perceptual RGB or linear RGB. Numbers from 0-255 (8 bits) or from 0-1 (math). In any case, due to the 3 orthogonal variables, we can model it as a 3 dimensional 'space', the color space, which goes from 0 to 1 on all 3 axes, which makes a cube, the color cube. Awesome. We have a cube. Every printable color can be put int his cube. Alsmost. Fluorescent colors and phosphorescent colors can actually produce more green or purple than shines on it, for instance, therefore seeming to have >100% reflection. Oh well, let's forget about them.
So, we have these 3 numbers, R, G and B. They are a point inside the color cube. Let's say we want to do color separation on this thing (which we do, that's why we're here). We can simply look at the 3D cube, which is a thought, as there's no color cubes anywhere. They are impossible, physically. But we can think about them in that way. Now, if we want to separate a color into other colors, selected from a set (out inks), then we need to first decide which colors we'll use to make up this color. Essentially, what we're doing, is we're carving up the color cube into pieces. Each piece has Only colors in it that can be mixed with a certain subset of our inks. We can have a chunk of cube that will be mixed with red, white, yellow, and black ink. Perhaps we'll use it to make skin colors. We can have another chunk full of colors that can be mixed with green, blue, white and black. That could be the sky.
Note that what was separated here was not at all the image. Instead, it was the color space. Which color space? Well, the color space in which the image is represented. That'll probably be the sRGB color space. We chopped up this space into little chunks, each representing a set of inks we can use to mix those colors.
The next step would be to have some kind of formula, a function, really, that takes the description of a chunk and a color that we know is in that chunk, and then spits out the amounts that the inks of this chunk will need to be mixed in to make that color we want.
Once we have that, it becomes trivial to pump the colors through this system:
1) Determine the chunk
2) Determine the ink amounts for each ink in that chunk
3) Do some adjusting to account for ink order and some physical ink properties
And what we did is, we pushed the image into the system, and the system truly separated the image. This is diametrically opposed to pulling colors. With a pull, the seps are getting pulled from the image, instead of the image 'pushing' onto the seps. The main reason it's so very different is that the one pull cannot 'communicate' with the next pull. It cannot alert the pink pull that the red was already pulled. The pink will therefore re-pull the red.
In the chunk system, it's the chunker that decides what chunk a color should be in and the formula, which decides for the entire color what inks should be used, all at once. So all separations are done simultaneously, and not 'by the seps', but 'by the formula'.
Whether or not you follow my ramblings or even if you are interested, there's a philosophy behind it, which is nice to know.
I set out to take this idea of pure math precision, math with a real base for existing, based on vector math in the 3D cube, and a nifty cube carving algorithm, and build it into a real system. This has not been easy. The math is downright painful.
Nonetheless, I've managed to do it. I can therefore guarantee that before adjustments, the color mixes are absolutely accurate to 12 digits beyond the decimal point. Note that the final films have less than 3 digits (0-255 for grayscale seps). Most sep systems kind of guess at these numbers. A lot of trial and error went into it, and it has produced some amazing award winning shirts. But now it is time to no longer rely on great guessing, since now we have an absolute accurate system. All that remains is to make it extremely easy to use and to sharpen the physical model. That way we can translate back from mathelonian into highly accurate physical shirts. I think I've already done a great job, but we do need some more real prints to make adjustments.
In the mean time, if you need a sep,
go here and get it for free. Highly accurate and no guarantees
. Over 60,000 lines of code I wrote myself. Over 4 years of work. This is not a gimmick, this is the real deal. I didn't stumble upon it, I didn't take it from anyone, I built it from the ground up with one purpose: to produce impeccable seps.
That said, if you have complaints, I'll see how I can improve the system. All you need to do is voice them. But note that there are currently a lot of features missing, so the bulk of complaints probably relate to these missing features.
Boy, this post is very long.
