WEBVTT Kind: captions Language: en 00:00:00.080 --> 00:00:01.640 What color is this? 00:00:01.640 --> 00:00:03.660 You’re probably thinking, "it’s red!" 00:00:03.660 --> 00:00:05.630 which, well, it is. 00:00:05.630 --> 00:00:06.630 And what about this? 00:00:06.630 --> 00:00:08.890 Why, it’s green, of course! 00:00:08.890 --> 00:00:13.570 And what video on color would be complete without an appearance by our old friend blue? 00:00:13.570 --> 00:00:18.449 We’ve been using these three colors to fool our eyes and brains into thinking that we’re 00:00:18.449 --> 00:00:21.390 looking at a full-color image for over a century. 00:00:21.390 --> 00:00:24.680 We can do this because of how our eyes and brains perceive color. 00:00:24.680 --> 00:00:29.320 It’s all about ratios, and though it often seems a little freaky, we can mimic the effect 00:00:29.320 --> 00:00:33.620 of any real color using just these three in controlled amounts. 00:00:33.620 --> 00:00:37.300 But now that we have the luxury of bringing the primary colors of light into the real 00:00:37.300 --> 00:00:42.630 world with bright, monochromatic LEDs, we can get a glimpse into just how wonderfully 00:00:42.630 --> 00:00:46.160 strange our sense of color perception actually is. 00:00:46.160 --> 00:00:49.969 In this video, we’re going to look at a series of demonstrations where objects in 00:00:49.969 --> 00:00:53.570 the real world are lit using light from the digital world. 00:00:53.570 --> 00:00:58.350 What we’ll find is that things can behave a little… unexpectedly when we play around 00:00:58.350 --> 00:00:59.350 with light. 00:00:59.350 --> 00:01:01.449 None of the footage in this video has been altered. 00:01:01.449 --> 00:01:06.039 I promise no matter how weird some of this looks, I’m seeing the same things in person. 00:01:06.039 --> 00:01:09.149 Let’s start with a brief overview of what it is we’re doing here. 00:01:09.149 --> 00:01:12.609 I’m using these RGB studio lights to provide illumination. 00:01:12.609 --> 00:01:16.850 I can control the ratio of red, green, and blue light they produce by adjusting their hue 00:01:16.850 --> 00:01:18.609 and saturation parameters. 00:01:18.609 --> 00:01:22.619 For the most part, we’ll be staying with a saturation of 100%, and this means we’ll 00:01:22.619 --> 00:01:27.159 be cycling through the 3 primary colors, red, green, and blue, as well as various shades 00:01:27.159 --> 00:01:31.520 of the three secondary colors that lie between them, yellow, cyan, and magenta. 00:01:31.520 --> 00:01:35.280 When I need to, I can switch to standard phosphor-coated white LEDs 00:01:35.280 --> 00:01:37.079 which provide a reasonable approximation 00:01:37.079 --> 00:01:39.149 of true, full-spectrum lighting. 00:01:39.149 --> 00:01:42.529 Yes, these lights really are G Bee’s knees. 00:01:42.529 --> 00:01:47.209 In RGB mode, the light they produce is trichromatic, just like our vision, but each individual 00:01:47.209 --> 00:01:51.679 color is monochromatic, meaning it’s comprised of a single wavelength. 00:01:51.679 --> 00:01:55.389 And this is where the breakdown between the real and digital world can occur. 00:01:55.389 --> 00:01:59.909 I’ll explain this in a little more detail shortly, but first let’s move on to a demonstration. 00:01:59.909 --> 00:02:03.439 We’ll be spending much of this video, in the dark. 00:02:03.439 --> 00:02:06.719 Here we have a kind of disappearing, color-changing ink. 00:02:06.719 --> 00:02:11.000 This whiteboard, when lit with apparently yellow light, appears to have some red writing 00:02:11.000 --> 00:02:12.000 on it. 00:02:12.000 --> 00:02:13.769 Well, watch this. 00:02:13.769 --> 00:02:15.410 Now it’s gone. 00:02:15.410 --> 00:02:20.900 But it re-appears, now as a slightly more orange color, with the presence of some blue light. 00:02:20.900 --> 00:02:24.110 Now watch as before your eyes the ink becomes a jet black. 00:02:24.110 --> 00:02:28.480 It stays black even as the light grows brighter and we approach cyan, before the black turns 00:02:28.480 --> 00:02:30.129 to red once more. 00:02:30.129 --> 00:02:32.030 And finally, it’s gone again. 00:02:32.030 --> 00:02:33.490 What’s happening here? 00:02:33.490 --> 00:02:36.629 Well, the ink on this whiteboard is in fact red. 00:02:36.629 --> 00:02:38.790 Switching to normal white lighting reveals that. 00:02:38.790 --> 00:02:43.231 The red ink absorbs nearly all of the light coming from the green and blue LEDs, which 00:02:43.231 --> 00:02:47.599 is why the ink appears black when the scene is anywhere between blue and green. 00:02:47.599 --> 00:02:50.709 It doesn’t reflect any of that light back into the camera. 00:02:50.709 --> 00:02:55.099 But in addition to absorbing the green and blue light, this red happens to be a near-perfect 00:02:55.099 --> 00:02:58.290 match to the red produced by the light’s red LEDs. 00:02:58.290 --> 00:03:00.670 And that’s why it disappears under red light. 00:03:00.670 --> 00:03:04.769 The white of the whiteboard reflects pretty much all of the red light back to the camera, 00:03:04.769 --> 00:03:08.159 as do most white objects, but so does the red ink. 00:03:08.159 --> 00:03:12.319 And so, there’s very little contrast between the ink and the board, and the ink effectively 00:03:12.319 --> 00:03:13.319 disappears. 00:03:13.319 --> 00:03:14.439 Let’s move on. 00:03:14.440 --> 00:03:16.740 What color is this can of spray paint? 00:03:16.740 --> 00:03:18.930 It’s pretty hard to tell, isn’t it? 00:03:18.930 --> 00:03:21.439 In fact, it’s impossible to tell. 00:03:21.439 --> 00:03:26.610 Right now, this can of spray paint is being lit solely by the red LEDs, which means it’s 00:03:26.610 --> 00:03:29.150 lit by a monochromatic light source. 00:03:29.150 --> 00:03:34.290 Doing this fundamentally breaks our color vision because we rely on the mixing of colors 00:03:34.290 --> 00:03:36.510 to determine what it is we’re seeing. 00:03:36.510 --> 00:03:39.709 Under the same red light, let’s look at some construction paper. 00:03:39.709 --> 00:03:41.999 This packaging says there are 8 colors here. 00:03:41.999 --> 00:03:44.319 Well, what on Earth are they? 00:03:44.319 --> 00:03:49.600 As far as I can tell, these are red, a darker red, a differently darker red, and uh, 00:03:49.600 --> 00:03:50.440 more red. 00:03:50.440 --> 00:03:53.600 I think there’s black, too, but I’m not sure. 00:03:53.609 --> 00:03:57.520 With only one wavelength of light available in this scenario, there’s just no way to 00:03:57.520 --> 00:03:59.450 know what it is you’re seeing. 00:03:59.450 --> 00:04:02.950 Notice how we cannot tell what the colors are on these Rubik’s Cubes. 00:04:02.950 --> 00:04:06.730 We can see that each color reflects the light back in different amounts, causing the stickers 00:04:06.730 --> 00:04:10.180 to appear in different brightness levels, but they’re all just different shades of 00:04:10.180 --> 00:04:11.459 the same red. 00:04:11.459 --> 00:04:16.070 But, with this being a Rubik’s Cube, we know the colors are white, yellow, orange, 00:04:16.070 --> 00:04:17.930 red, green, and blue. 00:04:17.930 --> 00:04:21.459 We can make some educated guesses into which stickers are which colors. 00:04:21.459 --> 00:04:25.930 The brightest are probably red, yellow, orange, and white, as these will reflect most or all 00:04:25.930 --> 00:04:28.220 of the red light back into the camera. 00:04:28.220 --> 00:04:30.400 The darkest are going to be blue and green. 00:04:30.400 --> 00:04:34.150 Now we can be reasonably sure the darkest of them all is blue, as that’s farthest 00:04:34.150 --> 00:04:36.900 from red, and the next darkest is green. 00:04:36.900 --> 00:04:38.419 But as far as the bright colors? 00:04:38.419 --> 00:04:40.300 That’s really anybody’s guess. 00:04:40.300 --> 00:04:44.710 The brightest is probably white, but then again there look to be too many that we might 00:04:44.710 --> 00:04:45.860 call white. 00:04:45.860 --> 00:04:49.819 So white and at least one other color look kinda the same. 00:04:49.819 --> 00:04:51.220 But which colors are they? 00:04:51.220 --> 00:04:53.880 Well, let’s switch the light over to white and find out. 00:04:53.880 --> 00:04:57.260 Oh, sorry, this one is actually monochromatic lemme, lemme get that out of here. 00:04:57.260 --> 00:05:01.639 So, we were right about green and blue, but orange, white, and yellow all appear to be 00:05:01.639 --> 00:05:02.870 the same. 00:05:02.870 --> 00:05:06.289 Red was actually slightly darker, which you might not have expected given that we were 00:05:06.289 --> 00:05:07.930 using red light. 00:05:07.930 --> 00:05:12.380 This tells us that the hue of this red is actually not purely red, as it does absorb 00:05:12.380 --> 00:05:14.599 some of the red we were throwing at it. 00:05:14.599 --> 00:05:18.610 And if yellow and orange were reflecting the same amount of red light back as white, well 00:05:18.610 --> 00:05:23.550 that again goes to show how strange our color perception is, and why monochromatic light 00:05:23.550 --> 00:05:24.680 breaks it. 00:05:24.680 --> 00:05:26.530 So how do we see in color? 00:05:26.530 --> 00:05:30.430 Well, in our eyes, we don’t just have a bunch of plain photoreceptors. 00:05:30.430 --> 00:05:35.259 We have some, known as rods, which just detect brightness, but those of us with typical trichromatic 00:05:35.259 --> 00:05:40.110 vision also have three types of color-sensitive cells, called cone cells. 00:05:40.110 --> 00:05:43.220 These are pigmented to filter the wavelengths of light that hit them. 00:05:43.220 --> 00:05:47.870 Now, we often think of these cone cells as being sensitive to red, green, and blue light. 00:05:47.870 --> 00:05:52.789 Which is broadly true, but their actual stimulation curves look like this. 00:05:52.789 --> 00:05:57.729 Notice how the medium and long cones, which correspond to green and red, kinda, overlap 00:05:57.729 --> 00:06:00.349 a lot, but the short cone is way over there. 00:06:00.349 --> 00:06:04.340 Well, where they are along the spectrum doesn’t actually matter all that much. 00:06:04.340 --> 00:06:08.220 What matters is that they respond differently to any given color. 00:06:08.220 --> 00:06:10.919 Say we have a yellow-green wavelength right here. 00:06:10.919 --> 00:06:16.159 Well, for this one color, and this one color only, the long and medium cones get equal 00:06:16.159 --> 00:06:19.629 stimulation, and the blue cones get negligible stimulation. 00:06:19.629 --> 00:06:24.069 This unique ratio allows our brains to interpret this color as yellow-green. 00:06:24.069 --> 00:06:28.760 As we move towards red and head into yellow, now the medium cone gets progressively less 00:06:28.760 --> 00:06:31.460 stimulated, and the long cone gets more stimulation. 00:06:31.470 --> 00:06:35.910 So, our brains know this color is closer to red than it is green. 00:06:35.910 --> 00:06:40.069 As we continue moving deeper into true red, the stimulation from the long cone starts 00:06:40.069 --> 00:06:44.139 to taper off, but the medium cone is tapering off faster. 00:06:44.139 --> 00:06:48.160 The important thing to remember is that any color at all along the visible spectrum will 00:06:48.160 --> 00:06:52.650 cause a unique ratio of stimulation between these three cells, and so our brains know 00:06:52.650 --> 00:06:54.530 what color that is. 00:06:54.530 --> 00:06:58.890 And so, we can easily fool our eyes and brains into thinking we’re seeing any color at 00:06:58.890 --> 00:07:01.990 all by using just three primary colors. 00:07:01.990 --> 00:07:07.030 We need one of them to be way over here, so that the long cone gets a fair bit of stimulation, 00:07:07.030 --> 00:07:09.229 but the medium cone doesn’t get all that much. 00:07:09.229 --> 00:07:10.530 So we’ll use red. 00:07:10.530 --> 00:07:15.250 We also need one to the left of the long-medium crossover, that way it stimulates the medium 00:07:15.250 --> 00:07:17.500 cone more intensely than the long. 00:07:17.500 --> 00:07:19.000 So we’ll use green. 00:07:19.000 --> 00:07:23.340 And of course, we also need one way over here that stimulates the short cone a lot, but 00:07:23.340 --> 00:07:25.160 doesn’t really influence the other two. 00:07:25.160 --> 00:07:26.430 So we’ll use blue. 00:07:26.430 --> 00:07:30.949 Now, to make a color like yellow-orange, we can simply mix red and green together, so 00:07:30.949 --> 00:07:33.770 that there’s a lot of red and a bit of green. 00:07:33.770 --> 00:07:39.060 This mixture causes the same stimulation that an actual yellow-orange object would. 00:07:39.060 --> 00:07:43.450 Because there’s overlap between the three cone cells, all real colors just cause a unique 00:07:43.450 --> 00:07:46.039 mix of stimulation between the three of them. 00:07:46.039 --> 00:07:50.110 That includes, by the way, white, which is all three in close to equal amounts. 00:07:50.110 --> 00:07:54.939 So, if we use three pure colors that allow us to selectively stimulate the three cells 00:07:54.939 --> 00:07:59.729 with any given ratio, we can artificially reproduce all visible colors. 00:07:59.729 --> 00:08:03.170 Our eyes simply don’t have a way to know they’re being fooled. 00:08:03.170 --> 00:08:07.340 But while we can make any color appear by using just three colors in different ratios, 00:08:07.340 --> 00:08:11.349 that doesn't mean that the world will look right without the whole spectrum to paint 00:08:11.349 --> 00:08:12.780 the whole picture. 00:08:12.780 --> 00:08:16.930 And unless we have a way to make the cone cells get stimulated in different ratios, 00:08:16.930 --> 00:08:18.910 we can’t see color at all. 00:08:18.910 --> 00:08:22.420 And with that in mind, let’s move onto some more demonstrations. 00:08:22.420 --> 00:08:24.720 This scene contains many red objects. 00:08:24.729 --> 00:08:28.750 But, under monochromatic blue light, you’d never know. 00:08:28.750 --> 00:08:32.960 Watch what happens, though, when I add just the tiniest amount of red light. 00:08:32.960 --> 00:08:36.180 Suddenly, the red pops into existence. 00:08:36.180 --> 00:08:39.950 This is a pretty trippy effect in person, because it’s as if someone’s messing with 00:08:39.950 --> 00:08:42.700 the RGB sliders of real life. 00:08:42.700 --> 00:08:48.620 Until we have red light available, red objects appear, well, grey or black. 00:08:48.620 --> 00:08:51.090 Even with green light, the same thing occurs. 00:08:51.090 --> 00:08:55.160 Notice how with green and blue light together, we can start to see the yellow and oranges 00:08:55.160 --> 00:08:58.070 of the Rubik’s Cube become distinct from the blue and green. 00:08:58.070 --> 00:09:01.970 Still, though, the red objects remain completely dull. 00:09:01.970 --> 00:09:04.810 Pure green light keeps them in the dark, just like blue. 00:09:04.810 --> 00:09:09.390 Keep in mind that the green light is still stimulating the long cones a fair bit, but 00:09:09.390 --> 00:09:14.130 without a third, longer wavelength to allow for comparison between the long and medium 00:09:14.130 --> 00:09:16.710 cones, our brains cannot see red. 00:09:16.710 --> 00:09:22.400 Plus, since the red objects in the scene aren’t reflecting any of that green light, they stay dull. 00:09:22.400 --> 00:09:26.520 Add just a hint of red, though, and suddenly the scene explode into color. 00:09:26.520 --> 00:09:31.250 Now, there is red light to be reflected, and more importantly for our eyes, there is red 00:09:31.250 --> 00:09:34.490 light to be detected and compared with green and blue. 00:09:34.490 --> 00:09:36.260 Here’s a different kind of color. 00:09:36.260 --> 00:09:38.090 A game boy color. 00:09:38.090 --> 00:09:41.660 Under blue light, this thing looks weird to say the least. 00:09:41.660 --> 00:09:44.790 Now I’ll add a bit of red and green, alternately. 00:09:44.790 --> 00:09:48.280 Compare the light on my hand to the light on the game boy, and you’ll see that overall, 00:09:48.280 --> 00:09:51.420 I’m not changing the color in the scene much at all. 00:09:51.420 --> 00:09:53.910 But the game boy drastically changes. 00:09:53.910 --> 00:09:56.830 This game boy’s color, by the way, is dandelion. 00:09:56.830 --> 00:10:00.440 Which is of course, to our eyes, a mixture of red and green. 00:10:00.440 --> 00:10:05.170 An important thing to note is that, just like our eyes, the camera’s Bayer filter (which 00:10:05.170 --> 00:10:09.550 actually separates subpixels into red, green, and blue elements) doesn’t filter red, green, 00:10:09.550 --> 00:10:10.740 and blue perfectly. 00:10:10.740 --> 00:10:12.620 There’s a lot of overlap. 00:10:12.620 --> 00:10:16.660 And I can show it to you, even with only one wavelength to see. 00:10:16.660 --> 00:10:21.580 You might assume that if light is a monochromatic green, then the camera’s blue and red subpixels 00:10:21.580 --> 00:10:23.130 will never become active. 00:10:23.130 --> 00:10:24.920 But this isn’t true. 00:10:24.920 --> 00:10:28.960 If I overexpose the image, you’ll see that it starts turning white. 00:10:28.960 --> 00:10:33.180 That happens because even though the light source is monochromatic, the red and blue 00:10:33.180 --> 00:10:38.410 filters will still let some through, so the image starts to turn white with enough exposure. 00:10:38.410 --> 00:10:40.590 The same thing happens with blue and red. 00:10:40.590 --> 00:10:43.680 However, this doesn’t mean we can start to tell colors apart. 00:10:43.680 --> 00:10:48.460 We still only have one wavelength illuminating the scene, which means the ratio of stimulation 00:10:48.460 --> 00:10:51.110 in the camera’s subpixels stays the same. 00:10:51.110 --> 00:10:54.410 The camera’s method of vision is surprisingly similar to our eye’s. 00:10:54.410 --> 00:10:57.050 Well, as a matter of fact it’s built for our eyes. 00:10:57.050 --> 00:11:01.560 And even under normal exposure levels, there is some green slipping in. 00:11:01.560 --> 00:11:06.130 You might expect the image to turn black if I remove all of the red channel, but in fact 00:11:06.130 --> 00:11:09.130 there’s a faint green image hiding underneath. 00:11:09.130 --> 00:11:14.470 That green is actually helping to define the ultimate hue of the red we’re seeing on-screen. 00:11:14.470 --> 00:11:16.770 Which brings me to my next demonstration. 00:11:16.770 --> 00:11:21.830 We can have a monochromatic light source of any color, not just red, green, and blue. 00:11:21.830 --> 00:11:26.290 With RGB lights, I can only produce yellow light by mixing red and green. 00:11:26.290 --> 00:11:30.671 This then becomes a dichromatic color, and if I illuminate this scene with it, we can 00:11:30.671 --> 00:11:33.030 actually tell some of the colors apart. 00:11:33.030 --> 00:11:35.430 We can even kinda tell blue from green. 00:11:35.430 --> 00:11:40.980 But, if I break out my yellow traffic light module (or amber, whatever), this is in fact 00:11:40.980 --> 00:11:42.940 a monochromatic yellow. 00:11:42.940 --> 00:11:47.370 This color looks quite similar to the yellow I’ve been making by mixing red and green, 00:11:47.370 --> 00:11:49.300 but it’s actually very different. 00:11:49.300 --> 00:11:53.380 So now, even though the green and red subpixels are both getting stimulation from the yellow 00:11:53.380 --> 00:11:58.910 light, because it’s actually just yellow they always receive the same relative stimulation 00:11:58.910 --> 00:12:00.670 no matter what’s in the scene. 00:12:00.670 --> 00:12:05.780 Our eyes, and the camera, both see these two sources of light as essentially the same color, 00:12:05.780 --> 00:12:09.630 but if we use them to illuminate the real world, and take a look at how they get reflected 00:12:09.630 --> 00:12:12.990 back, we discover they’re actually very different. 00:12:12.990 --> 00:12:16.410 And that brings us to what makes this whole ordeal so messy. 00:12:16.410 --> 00:12:20.700 You may have heard of a term called the color rendering index, or CRI. 00:12:20.700 --> 00:12:24.690 This describes how well an artificial light source reproduces the color of the objects 00:12:24.690 --> 00:12:25.770 around us. 00:12:25.770 --> 00:12:31.551 Incandescent lights, being a blackbody radiator, had a perfect CRI, just like the sun, but 00:12:31.551 --> 00:12:36.290 more efficient LED and fluorescent light sources, indeed practically all light sources that 00:12:36.290 --> 00:12:40.390 aren’t incandescent, don’t emit light as a perfectly uniform spectrum. 00:12:40.390 --> 00:12:45.000 Now, as we know, one of the most common ways to mimic white light is to produce red, green, 00:12:45.000 --> 00:12:48.490 and blue light, because, well, if you haven’t figured that out by now you’ve not been 00:12:48.490 --> 00:12:49.740 paying much attention. 00:12:49.740 --> 00:12:53.920 This works absolutely fantastically for creating a display device like the one you’re staring 00:12:53.920 --> 00:12:55.150 at now. 00:12:55.150 --> 00:12:59.200 Because it’s providing its own illumination, it doesn’t need to worry about how the red, 00:12:59.200 --> 00:13:02.750 green, and blue channels interact with the objects around you. 00:13:02.750 --> 00:13:06.530 It just needs to fool your eyes into thinking they’re looking at a full-color image. 00:13:06.530 --> 00:13:11.421 And, well, displays are getting better and better, with incredibly lifelike colors, all 00:13:11.421 --> 00:13:13.960 from just three colors of light. 00:13:13.960 --> 00:13:18.160 Except for that one time Sharp got all weird with the yellow subpixel which was absolutely 00:13:18.160 --> 00:13:23.070 unnecessary especially since nobody’s encoding color in an RGB-Y space, but I digress. 00:13:23.070 --> 00:13:27.020 But the problem with using just three colors of light to illuminate the real world is that 00:13:27.020 --> 00:13:29.080 this rarely looks right. 00:13:29.080 --> 00:13:30.930 Think about that whiteboard earlier. 00:13:30.930 --> 00:13:33.510 The red ink was invisible under red light. 00:13:33.510 --> 00:13:37.110 This meant that it reflected practically all of the red light back. 00:13:37.110 --> 00:13:41.750 Now, imagine I’m using these lights with red, green, and blue all working together. 00:13:41.750 --> 00:13:43.320 This looks white to my eyes, 00:13:43.320 --> 00:13:46.120 but when it gets reflected off of the objects around me, the 00:13:46.120 --> 00:13:49.360 ratio of colors coming back can be way off. 00:13:49.360 --> 00:13:53.290 In the case of the whiteboard, the red looks way too intense and bright. 00:13:53.290 --> 00:13:54.710 Which makes sense. 00:13:54.710 --> 00:13:59.360 If one third of the light from these lights is red, and the red ink reflects all of it, 00:13:59.360 --> 00:14:05.600 it’s suddenly freakishly bright because, well, red is not one third of the color spectrum. 00:14:05.600 --> 00:14:09.780 Under true white light, a much greater percentage of light gets absorbed, and the red appears 00:14:09.780 --> 00:14:11.710 more dull, like it should. 00:14:11.710 --> 00:14:15.810 As a quick note, this is the one demonstration where the camera couldn’t quite capture 00:14:15.810 --> 00:14:17.560 what my eyes were seeing. 00:14:17.560 --> 00:14:20.210 The difference in person is much more dramatic. 00:14:20.210 --> 00:14:24.600 The problem here is that the ability to reduce the real world into three wavelengths of light 00:14:24.600 --> 00:14:26.470 is not reversible. 00:14:26.470 --> 00:14:31.090 If we have a truly white light source, then all the in-between colors get reflected as 00:14:31.090 --> 00:14:32.620 they truly are. 00:14:32.620 --> 00:14:37.260 Our eyes can see any wavelength of light because of all that overlap between the cone cells. 00:14:37.260 --> 00:14:41.350 And indeed, cameras can see any wavelength of light, because their RGB bayer filters 00:14:41.350 --> 00:14:43.570 also have overlap between them. 00:14:43.570 --> 00:14:48.080 And so we can reproduce the stimulation real objects cause in our eyes with just three 00:14:48.080 --> 00:14:52.360 wavelengths of light, but we cannot expect those three wavelengths to produce the same 00:14:52.360 --> 00:14:57.290 stimulation ratios that they should when they hit and get reflected off of real objects 00:14:57.290 --> 00:14:58.550 in the real world. 00:14:58.550 --> 00:15:01.800 This can perhaps best be demonstrated by the color purple. 00:15:01.800 --> 00:15:04.440 Purple is a rather strange color in general. 00:15:04.440 --> 00:15:08.850 It, along with magenta, are what are called non-spectral colors. 00:15:08.850 --> 00:15:13.240 If you look on the color spectrum, you’ll find violet just on the other side of blue, 00:15:13.240 --> 00:15:17.940 but true violet is rather dull, and in fact we have a hard time seeing it. 00:15:17.940 --> 00:15:22.610 Which is no surprise since it barely registers with any of our cone cells. 00:15:22.610 --> 00:15:28.110 Purple and magenta are kinda similar to real violet, but in a sense, these colors exist 00:15:28.110 --> 00:15:29.960 only in our minds. 00:15:29.960 --> 00:15:32.490 That’s pretty wild, when you think about it. 00:15:32.490 --> 00:15:37.160 Now obviously purple things exist in nature and we can see them with our eyes, so it’s 00:15:37.160 --> 00:15:39.060 not like the color is imaginary. 00:15:39.060 --> 00:15:42.980 But, it cannot be reproduced with a single wavelength of light. 00:15:42.980 --> 00:15:48.280 We only see purple and magenta when our eyes receive blue and red stimulation, 00:15:48.280 --> 00:15:50.040 but little green. 00:15:50.040 --> 00:15:54.710 Therefore, purple and magenta objects absorb a fair bit of green light, but reflect both 00:15:54.710 --> 00:15:56.330 red and blue. 00:15:56.330 --> 00:16:01.310 And luckily, our brains have synthesized this combination of stimulation into magenta, and 00:16:01.310 --> 00:16:05.440 not the average wavelength between them, as we do with yellow and cyan. 00:16:05.440 --> 00:16:07.640 Otherwise, it would be another green. 00:16:07.640 --> 00:16:10.760 Anyway, let’s take a look at our old friend Putt-Putt. 00:16:10.760 --> 00:16:15.360 This particular anthropomorphic automobile is a rather vibrant shade of purple. 00:16:15.360 --> 00:16:19.240 Now, using the phosphor-coated white LEDs, he looks pretty normal. 00:16:19.240 --> 00:16:22.060 But when I switch to the RGB LEDs, 00:16:22.060 --> 00:16:24.130 well not so much. 00:16:24.130 --> 00:16:26.230 Under green light, he looks pretty dull. 00:16:26.230 --> 00:16:30.760 Which we might expect, given that we can of course make purple by mixing red and blue 00:16:30.760 --> 00:16:34.940 pigments, which will together absorb mostly green wavelengths. 00:16:34.940 --> 00:16:39.040 When we add blue light, well now he just looks blue. 00:16:39.040 --> 00:16:44.710 All into the cyan range, Putt-Putt looks just like a blue, and once we hit blue, well now 00:16:44.710 --> 00:16:50.280 he looks kinda like a grey, as his white features become blue, and his body becomes a slightly 00:16:50.280 --> 00:16:51.750 darker blue. 00:16:51.750 --> 00:16:53.330 But here’s the weirder thing. 00:16:53.330 --> 00:16:56.250 Add red, and now he really looks grey. 00:16:56.250 --> 00:17:01.760 If I change the angle so you can see his tongue, yes cars have tongues, duh, his tongue is 00:17:01.760 --> 00:17:05.280 bright red, but his body still looks grey. 00:17:05.280 --> 00:17:10.449 And perhaps stranger still, replace the blue with some green and move into yellow territory 00:17:10.449 --> 00:17:12.400 and he looks… burgundy? 00:17:12.400 --> 00:17:14.020 A burnt red? 00:17:14.060 --> 00:17:17.360 I don’t know exactly what this color is, but it is not purple. 00:17:17.360 --> 00:17:22.000 Now, some of this is down to how our brains’ white balance works, as we are comparing his 00:17:22.000 --> 00:17:26.819 white eyes to his body color, and in fact if we look in Photoshop we’ll see that what 00:17:26.819 --> 00:17:29.429 looked grey to us is actually fairly purple. 00:17:29.429 --> 00:17:32.820 It’s not the right purple, but it is purple. 00:17:32.820 --> 00:17:35.400 And when you think about it, that makes perfect sense. 00:17:35.400 --> 00:17:40.920 Assuming this shade of purple is just a darker magenta, then if lit with magenta light, his 00:17:40.920 --> 00:17:46.379 body would appear to be the same hue as his white features, but at a reduced intensity. 00:17:46.379 --> 00:17:51.940 Without any sort of color contrast, that reduction in intensity just looks … grey. 00:17:51.940 --> 00:17:56.920 Grey is simply a darker version of white, and what is white in this scene, is actually 00:17:56.920 --> 00:17:58.200 magenta. 00:17:58.200 --> 00:18:01.500 This also explains why his tongue looks so vibrant. 00:18:01.500 --> 00:18:06.049 His tongue is now the only thing actually changing the relative amounts of color being 00:18:06.049 --> 00:18:07.690 reflected back. 00:18:07.690 --> 00:18:12.350 Since it absorbs blue like a good red should, it’s now able to set itself apart from the 00:18:12.350 --> 00:18:15.070 magenta mess that is everything else. 00:18:15.070 --> 00:18:18.740 And of course, we can also explain why he looks red under yellow light. 00:18:18.740 --> 00:18:22.370 His body will be absorbing most of the green coming from the lights, so the only thing 00:18:22.370 --> 00:18:24.519 it reflects back is red. 00:18:24.519 --> 00:18:28.659 It looks a little weird because of the fact that it does absorb some of the red just as 00:18:28.659 --> 00:18:32.470 it absorbs some blue, so it looks darker than his tongue. 00:18:32.470 --> 00:18:36.409 And our brains’ vain attempt to compensate for the yellow light and assume that’s real 00:18:36.409 --> 00:18:38.840 white makes it look stranger, still. 00:18:38.840 --> 00:18:41.289 Now we’re not quite yet done with Putt-Putt. 00:18:41.289 --> 00:18:45.299 So far, I’ve been showing you how he looks under various colors of light. 00:18:45.299 --> 00:18:50.580 But even under apparently white light, comprised of red, green, and blue, this purple color 00:18:50.580 --> 00:18:54.700 simply does not get rendered correctly at all. 00:18:54.700 --> 00:18:59.720 Notice how differently he looks under normal white light using the phosphor-coated LEDs, 00:18:59.720 --> 00:19:03.990 compared to the false white made by the RGB LEDs working together. 00:19:03.990 --> 00:19:08.450 Something about the way this purple absorbs wavelengths in the visible color spectrum 00:19:08.450 --> 00:19:12.919 simply cannot be reproduced using a trichromatic RGB light source. 00:19:12.919 --> 00:19:15.030 At least, not these lights. 00:19:15.030 --> 00:19:19.659 So keep in mind that even though I can show you this royal purple on a screen using only 00:19:19.659 --> 00:19:24.830 some red, some green, and some blue, I can’t just use those three colors in the real world 00:19:24.830 --> 00:19:27.110 and expect to achieve the same result. 00:19:27.110 --> 00:19:31.580 Now, before I leave you, well first of all that can of paint was yellow, sorry I forgot 00:19:31.580 --> 00:19:35.320 to answer that earlier, but more importantly while setting these demos up I think I may 00:19:35.320 --> 00:19:40.190 have accidentally discovered one of the most effective ways to understand color blindness. 00:19:40.190 --> 00:19:44.710 Now, I’ve seen lots of simulated images online, but they’ve never really clicked 00:19:44.710 --> 00:19:46.550 with me like this did. 00:19:46.550 --> 00:19:50.270 The most common type of color blindness is red-green colorblindness. 00:19:50.270 --> 00:19:55.389 There are varying degrees of this deficiency but in general it means that the green / medium 00:19:55.389 --> 00:19:57.920 cones are either malfunctioning or not present. 00:19:57.920 --> 00:20:02.999 Now, I have no way to turn down or otherwise stop the green cones in my eyes from working. 00:20:03.000 --> 00:20:08.720 But, if I light the room I’m in with dichromatic magenta light, the effect is somewhat similar. 00:20:08.720 --> 00:20:11.529 Now, it’s not like this is what a color-blind person sees. 00:20:11.529 --> 00:20:16.040 Especially because the entire scene is intensely colored, and green objects, like this marker, 00:20:16.040 --> 00:20:18.710 appear very dark, not simply similar to red. 00:20:18.710 --> 00:20:23.350 But, for the first time, I truly felt like I could not distinguish red and green all 00:20:23.350 --> 00:20:24.350 that well. 00:20:24.350 --> 00:20:29.139 The snake figure, here, suddenly had its red and lime green become awfully similar. 00:20:29.139 --> 00:20:33.250 Again, this is by no means accurate, look at how the green stickers on the Rubik’s 00:20:33.250 --> 00:20:37.429 Cube look black, but it is certainly interesting to have the color information of the real 00:20:37.429 --> 00:20:40.700 world become limited in ways I’ve never experienced. 00:20:40.700 --> 00:20:42.789 Anyway, that’s it for now, I think. 00:20:42.789 --> 00:20:46.630 I didn’t buy these lights assuming I was going to make a video about how strange RGB 00:20:46.630 --> 00:20:50.230 lighting is, but playing around with them led to some interesting places. 00:20:50.230 --> 00:20:55.000 And honestly, it’s helped me understand color vision even better than I did before. 00:20:55.000 --> 00:20:58.080 Thanks for watching, and as always a huge thank you goes out to the people supporting 00:20:58.080 --> 00:20:59.370 this channel on Patreon. 00:20:59.370 --> 00:21:03.720 Thanks to the support of people like you, I can make bizarre little detours like these, 00:21:03.720 --> 00:21:04.990 and I really enjoy it. 00:21:04.990 --> 00:21:06.320 I hope you do, too. 00:21:06.320 --> 00:21:10.360 If you’d like to join these people in supporting my work, you can check out the link in the description. 00:21:10.360 --> 00:21:12.880 Thanks for your consideration, and I'll see you next time! 00:21:13.920 --> 00:21:16.920 ♫ trichromatically smooth jazz ♫ 00:21:18.820 --> 00:21:19.320 Hey! 00:21:19.320 --> 00:21:20.160 It’s me! 00:21:20.160 --> 00:21:21.480 But from the future! 00:21:22.060 --> 00:21:22.560 Woah. 00:21:23.380 --> 00:21:27.340 So many of you probably know this but if you didn’t, I have a second channel where I 00:21:27.340 --> 00:21:31.889 sometimes upload rather random things, they tend to be kind of rambly, and I wanted to 00:21:31.889 --> 00:21:36.659 let you know that following this video I want to have a more relaxed discussion about some 00:21:36.659 --> 00:21:41.779 of the subtle differences between using true white lighting and RGB white lighting. 00:21:41.779 --> 00:21:45.350 So if you want to check that out, there’s gonna be a link in the description as well 00:21:45.350 --> 00:21:47.700 as a card on the end screen. 00:21:47.700 --> 00:21:51.740 For now, I hope you’re enjoying this rather groovy looking Rubik’s Cube. 00:21:51.740 --> 00:21:53.440 It's pretty groovy looking. 00:21:54.180 --> 00:21:54.900 Groovy.