The first 100 people to use code ACEROLA at the link below will get 60% off of Incogni: incogni.com/acerola ! #ad Back to basics with this one. Thanks for 100k subs! I will be making the 100k sub special in November, please look forward to it. Check out my patreon for the potential topics for next month though!
Honestly, if anything, Incogni is *more* insidious than the data brokers that it supposedly protects you from. Data brokers might profit off your personal information, but Incogni profits off your perceived anxieties over people profiting off your personal information. It's fearmongering as a business, just like VPNs. The reality is that you are not that important and that there is nobody specifically looking up your personal information to do bad things with it. Literally nobody. Stop freaking out about things that happen only in your imagination.
Exactly, like me who started decades ago and still try to make sure im understanding the concepts of this sh*t to this day and forced me to study pholosophy cause all i wanted was to actually understand it. it just drove me insane.
This video feels like a severe case of ADHD that managed to condense years of Art academy in 37 minutes while still talking about the same topic, it's glorious!
One of the best parts of acerola's videos to me is that they usually progress the same way my brain works. Start talking about how to create pleasing color palates at random and after a while is discussing why posters design aren't as creative and informative as they used to be, just to stop and think "way why did i get here in the first place?! Oh yeah random colors!"
13:55 : I’ve heard that mantis shrimp perception actually just uses “which of the 16 cone types is responding the most in this region”, so their color perception is, weirdly, just 16 colors, rather than a point in a 16-dimensional space.
I absolutely love when evolution does stupid shit like this. I mean, it's pointless; why not just have a bit of colour blending and get sooo much more visual acuity. but does it work, even if only just _well enough,_ at least? yes, apparently. and so it do be like that. (for those interested in other wonderful evolutional fuckery, go look up how giraffes have a nerve that goes from the brain, down to loop behind a big artery near the heart, and _up to the mouth again_ - this is the case for all mammals, but with giraffes it's particularly egregious. they have actual mouth-movement lag)
@@mnxs yup, the biggest problem this does for giraffes is that they simply can't make noises properly. their vocal cords can never operate in sync when half of it is suffering permanent ping issues.
As somebody who has gone down the digital colour hellhole over the last few years, this video is QUITE GOOD. You've done a great job at summarizing many important concepts and this is a resource I'll absolutely be sharing with folks who I need to introduce the topic to in the future. Really well done!
I agree! as a color science hobbyist this video covers a lot of information quite accurately. The only major mistakes that I noticed were the intuition for what X, Y, and Z change about a color and the definition of chromaticity. TLDR: XYZ is very similar to the RGB color model in that it is additive with the X 'primary' being reddish, the Y greenish, and the Z bluish. Chromaticity means a specific point on the chromaticity diagram i.e. color without regard to luminance. and here's the long version: XYZ is very similar to an RGB color model in that it is additive with the X 'primary' being reddish, the Y greenish, and the Z bluish. The catch is that the X and Z axes are perpendicular to the luminance axis which means that they would theoretically (being impossible colors) appear black on their own. Only by adding a little bit of luminance (Y) would you be able to get an idea of what they look like. You can also think of X as leading to the right in the chromaticity diagram, Y leading up, and Z leading to the lower left corner. Chromaticity means a specific point on the chromaticity diagram i.e. color without regard to luminance. Another way of thinking of chromaticity is to remember that the chromaticity diagram is a projection of a slice of the XYZ space. Imagine a ray through the origin of the XYZ or RGB color spaces. Technically the origin is black and is equivalent to any chromaticity at 0 luminance, but any other points on a ray will have the same chromaticity while differing only in luminance.
Also went down this rabbit hole a few years ago. I ended up on CIECAM02 which since OKLAB wasn’t really that popular back then. I wish this video existed back then because it would’ve saved me weeks of research.
In terms of doing math with colors (such as photoshop layers blending modes), do you believe it is better to work with a perceptual models such as Oklab or instead work with ACEScg or other industry-standard linear RGB spaces? I'm new to all this theory and I am interested in any insight about the differences between these
I can only agree with Hank. I still have to find a good tristimulus dataset to convert any pure wavelength to a perceptual color, in a way that when you draw a spectrum, it's not plagues with visible bands. Meanwhile I devised my own color model, HCI (Hue Chroma Intensity) that arguably works better than *Lab color models, but is computationally very expensive to invert (requires binary search).
@@Uncl3M3at Both have their value. It's about using the right tool for the job. At the end of the day, your output will be a formed RGB image, the steps to get there are of course entirely dependant on your specific situation. It's preferential to use OKLab to calculate gradients for the reasons explained in the video, but you're going to have to convert those generated values to RGB values so you can actually use them in a bitmap image or display them on a screen. As for the blend modes, those also have value as they are used for artistic effect and produce a (presumably) desired response. They're equally valid.
i suspect one major reason HSL palettes were still frequently picked is that it will sometimes "incorrectly" give you colors with more chromaticity. for a nice gradient, maybe you want one of those three parameters to remain constant, but for an artistic palette i would want at least one color that pops over the rest. if the OKLab generator was to occasionally provide a nice vibrant color along with the rest of the gradient, you might see even more people favoring it.
as another digital artist, this is an amazing insight! i was a little thrown off by how much i liked some of the hsl palettes, but this explains it as well as how much more consistent oklab was in comparison!
Honestly, if more math teachers teached using graphical examples of what you can do with it like you do, I think quite a bunch of people would be more interested in mathematics, amazing video as aways Acerola.
Honestly I still prefer the topics being introduced in classes the same way. I think the graphic examples and uses are more helpful after to help solidify the concepts. The problem with that is educators have basically no time to do both, leading to an unsatisfactory experience regardless.
Honestly, having each class of students being taught by individual teachers is inefficient. The main source of knowledge should be the best of prerecorded lessons and an on-site teacher is just there to help with questions that can't be googled quickly and to keep distractions down.
Most digital artists who have a more painterly art style usually start with a grayscale base just to establish the contrasting values, then we add color on top of that using adjustment layers that doesn't affect the values, only the hues. That way we keep the dynamic contrast and can still use whatever color sliders we have at hand. Coloring a grayscale base isnt a universal thing ofc, but it has it's roots from oil painting techniques where artists paint a sepia toned underpainting that serves the same purpose.
Interesting! I would imagine that using a perceptual color space to calculate the adjustment layers would really make a difference there. Changing the hue and saturation while keeping the HSL/HSV "luminance" or "value" number the same is going to change the perceived value, in some cases quite a bit.
It truly amazes me how you can take such a seemingly small problem and dig into it to explain everything that is actually involved in solving it, which is a lot, while still managing to keep it all connected and fascinating. As a computer science student who is also into graphic design and painting, you genuinely make me consider taking a professional step towards graphics programming, which is something I didn’t even know about before I stumbled upon your channel. Thank for the inspirational videos and keep doing it like you do !
My senior capstone project in college was about palette selection from still images. I don't know anything about shaders or color theory, but because I've your videos I have been inspired to dust it off and see if I can get it to compile again. Edit: after jumping through a lot of hoops to make JavaFX work again, I got it to compile. 6 years ago me was not a good programmer. The basic concept was to analyze an image and select a limited color pallet that would keep that average (mean) color of the entire image the same. This mostly produced extremely ugly images, but occasionally made some really neat effects.
so I'm also a bio student, and I seek info. The RGB color model means that the screens don't show all colors we can see, or are those colors outside it the so called imaginary colors
@@HoneyDoll894 Both, if you're talking about the sRGB triangle. There are colors outside of sRGB that we see normally and sRGB can't replicate them due to the LCD screen limitations. You could still use those colors if you were to paint them in real life. There are also imaginary colors that only our eyes can see and we couldn't ever replicate a paint for those colors because those imaginary colors only exist in our minds under certain contexts like cone fatiguing.
@@HoneyDoll894 It really doesn't matter much what the visible spectrum for us humans is. The physical ability of LEDs or CRTs has its own intensity graph, all of them nonlinear. This is adjusted by filters on the electronics level to map that to a more usable spectrum. More translation is applied (try to switch the profiles of your monitor). Then we have screen-space colors and all the color models you mentioned. Another level of filters, in your drawing/video program will help you to produce/adjust content for print media, for web or other forms of distribution, like movie theaters ... you get what I mean, its simply overwhelming:) Direct RGB to, lets call it, human eye-receptor-mapping would be non-linear, because of our evolutionary higher sensitivity shaped by evolution, e.g. the greatest light sensitivity for spectral colors in the yellow-green range (hey? when did you eat a nice meal of grass, the last times? hehehe). The same applies to an arbitrary, for example 0-255, 32bit integer value encoding colors and applying them as a voltage/phase-angle-modulation/etc to light emitting technology. Those aren't linear either. Any of those filters in this possible long and complicated chain (from your Image data through the monitor into your eyes) help to make the experience more natural and also standardized. Well, as I said, there are several techniques for different applications. Viewing is a whole different approach than printing media. Btw that all of that does make really sense, one has to CALIBRATE his monitor. Mostly no one (artists and professionals of course excluded here), no normal computer user has such a device or cares about that. Color receptance is highly individual and can even change with your personal mood or, your hormones or the time of the day (most people don't have a standardized studio light background either ... but our lovely sun:) ). So this can be a topic of fights to the blood (of some nerds?) and also the most inconsistent science of all (depends of course if you "eliminate" those pesky humans out of the equation, hehehe). I wish you both a wonderful time and big discoveries and exciting insights during your studies!:) (This text was written in a dark grey color. Hmm, I should clean my monitor ...)
Same, I didn't know how bad until I had a class where we were supposed to create different color relationships with color swatches and I had such a hard time with it.
i don't want to whine too much about one of the admittedly less debilitating disabilities one can have but i really do think people underestimate just how much colorblindness impacts the way you process, and in turn depict, the world. like holy shit guys did you know that Light? i sure fucking didn't.
@@northropi2027 No see, as an artist who can see in all color, I swear to God I had a dream once where I went colorblind and was absolutely devastated. Here's an abridged version (it's really long I'm sorry ;-;) : 9/17/23 Sun. This was mostly in the first person. I woke up from a coma to my mom feeding me cornmeal mush or something. Everything was a warmish blue, and was dark. Looking in the bathroom mirror, I saw that my left eye was missing. Everything remained blue as I went to ask mom what happened, and she reluctantly told me that I'd somehow cut it off before passing out and falling into the coma; the surgeons couldn't help much. With only one eye, I lost my depth perception, which was to be expected, though it was still upsetting. Everything then turned black and white and I thought (for some reason) that I needed both eyes to percieve color; this only further devastated me. I had an ipad or tablet on hand, and opened up to a color wheel. I tried imagining the colors overlaid on top of it, but it felt horrible. I searched up art tutorials which didn't require color as much (charcoal, graphite, micron pen, anatomy, shape, texture, etc), "but with the media so rich in color, every other tutorial regarding gorgeous sunsets, vibrant floral tones, and other color palettes were hard to sit through, and I skipped through all of them." The rest of the dream was kind of weird because one of my legs was replaced with a lion leg and there were some interesting shenanigans, a water park, and two siblings having a rap battle...? Then there was one section that went, "The walking was inconvenient for sure, *but not so bad as the vision loss.* I would see flashes of color, though now I'm not sure if I was just imagining it. When I would think of things, I'd think of them in color, but I was already starting to forget exactly what they looked like--which particular hue it was, how warm my mom's face was, which gradient went over the shadows," this was followed by a look in the mirror again revealing that a wrinkly little eye was growing back somehow. Inaccurate as the actual aspects of colorblindness were, I think it's safe to say that this dream was within the top 5 tiers of my list of nightmares. It was awful.
I've been doing digital art in photoshop for a while now, and I always KNEW that the included color picker was frustrating because of the inherent values of some hues (e.g. fully saturated+luminant yellow is still a light value than fully saturated+luminant blue). I knew what I wanted in an updated color picker but couldn't even think how to describe it, and OKLAB is super interesting. This video had me glued to my screen and was hilarious the entire time, and I feel my brain growing wrinkles. This was super interesting, thank you for this video.
You're such an excellent communicator! It's cool that we're coming at the same problems from different sides: my background is in art and I've been approaching computer graphics understanding perceptual color first, and now trying to learn how people have tried to map it to the tech we use to view it.
As soon as I realized we were about to dive into color spaces and color theory, I buckled my seatbelt and called my parents in case I didn't make it. Glad to see you didn't self destruct half way through, only your clone :)
1. I saw a pale magenta (like #ffa1e4) 2. As a digital artist who learned color/drawing almost entirely digitally, this is a really cool video to watch! Digitally picked colors are my favorite part of the process, and while I manage to be good with digital color it is at the cost of being able to do literally anything else intuitively. And yes, HSL is useful but it can be a nightmare in terms of picking colors that we actually perceive as brighter. For me I use a special color profile to check, alongside the old fashioned eye squint to determine value. Also, I'm out of luck.
i definitely saw a bright saturated magenta too, but i really enjoyed the massive nothing square in the middle of my vision from the white square around the circle. the square persisted for much longer
I felt so heard when you listed Color Theory as most complex. I tinker and obsessively toy with RGB hex values for a living. ❤ Massively useful banger vid as always.
At 16:41 I actually saw a very very light blue/cyan colour. But I also have a color blindness of two types with mid/low intensity (prota+deutera-nomalies) so that's what might cause such a difference from the expected outcome. I also really loved your video about mentioned topic. Keep the good work!
Photoshop is not the only alternative nowadays. A painting software like Krita do offers Lab as well. Krita also allows to have a grayscale luminance view aside the canvas allowing for direct corrections.
@@CrescentUmbreon yes L*a*b* is available at least since 4.0 so it's been a while. Coupled with the ability to display different views (like grayscale or mirrored) this is really easy to directly paint without a complex preliminary sketch. Though I don't know if there is some view to get the overall median color/luminance of an image, to check its balance.
That oklab gradient at 32:35 is sooo satisfying, it absolutely blows HSL hilariously uneven luminance out of the water what the heck I need to use this thing RIGHT NOW.
As a student in computer graphics and image analysis, your videos are always entertaining and informative but this one especially so; OKLAB is a game-changer.
THANK YOU! Finding a way to map Human Colour Perception to a screen's representation of it has been so utterly difficult. Not just that, even understanding that issue is so tough and only now after watching this video do I feel like I understand where my biggest issues with understanding colour have come from. This will help me so so much
I love color theory. There are so many strange things that happen. Like if you mix the primaries RGB, which are additive colors, with the secondaries CMY, which are subtractive, you'll find that they oscillate between a color that was 50% lighter than you expected, neutral, and 50% darker than expected. When you mix green and red light, you get yellow, because the two colors add light together. What do you get when you mix green and red pigment/paint though? Most people would say you get black, but you don't, you get a brown. But why should the color, the hue be any different from when we were mixing light? It shouldn't change the hue, it should just make it seem darker. That's when it hit me, I had just watched a video on how brown isn't a real color, it's just context. Any reddish orange to yellow color when darkened can be considered brown. So when we mix our red and green light we get yellow, and when we mix the red and green pigment, they are subtractive and darken the yellow! THAT'S why it looks brown! You DO get the same color when mixing pigments as you do mixing lights. It's just darker. If you mix two subtractive colors (ie: red and green), it will be darker, mix an additive with a subtractive (ie: yellow and red) it will expected brightness, mix additive with additive (ie: yellow and magenta) it will be brighter than expected. And green is HUGE. It's is so massive compared to the other colors. You try to mix a color halfway between green and yellow, and you get green lol. You try to mix a color halfway between green and blue and you get green. It's crazy how much further toward the adjacent colors you need to go from green to get those tertiary colors.
Combining pigments is not the same kind of colour theory you have to consider what is taking place there is chemistry, it depends entirely on how those things react as to what the result will be not simply what their apparent colour is.
@@iisthphirchemistry? no. you're implying that there is a chemical reaction happening between the components of the paints. that is not the case. it's just pigments getting _mixed,_ that's all.
It depends on how the pigments react, that would be a chemical reaction. It could be extreme or basically nothing it just depends on what is in them. These days I think most types of paints are made such that they react very little one of the reasons being so you can mix them with predictable results. There are others that are made to react with the surface they are painted on for instance. Historically though pigments were made from all sorts of strange things many of which were not as stable and so did react to some degree, with each other or simply by changing over time from oxidation. Sometimes it may be necessary to paint in one colour to have it end up being a different one. As, I understand, was the case with some of those used by the Italian's during the renaissance.
@@iisthphir it seems unwise to make a historical parallel to modern paints; the latter are often polymer-based or otherwise examples of modern chemical technique and considerations. if two paints are mixed, they are usually from the same manufacturer and of the same type, and thus have the same base chemical composition, _except_ the pigments - thus compatible. but sure, if you mixed a two paints of radically different compositions, weird stuff might happen. I still doubt it would attack the pigments themselves much, though. given modern environmental, toxicological and general safety requirements, I believe pigments wouldn't be particularly reactive anyway - not to mention that it's just not a very desirable trait for a supposedly stable pigment to have. also, the chemical reactions that might happen with the painted surface are one thing, and shouldn't affect the pigments by design. if the paints share a common base composition, and that base won't attack one pigment when it's by itself, why would it attack when it's mixed with something with a different pigment? besides, any non-trivial colour paint would most likely be a mix of any number of pigments anyway.
12:22 Fun fact: because of red’s lack of luminance red light does not cause your vision to switch to “cone mode”, meaning that when you turn off a red light your eyes with still be adjusted to the dark! This can be very useful at night.
I was so intruiged what the mystery behind those three pallette generators was! Thank you and I'm now going to have "OKLAB" in a drawer somewhere in my mind in case I ever need it!
This was a psychological horror movie. Anyway, amazing work, I can’t imagine the amount of research that goes into these videos, as a digital artist this was fascinating, and gave a nice explanation of the problems with HSL that I think we all mostly assumed were just something we’d have to live with, and also why they exist
A big problem I come across is how all my screens are not calibrated by default, and I assume most viewers viewing my art do not have professionally calibrated screens either. So when creating something, I usually tend to view it on different screens to make sure it looks kind of okay on all of them, but it's impossible to be too picky about this stuff. Maybe I should actually get my screen calibrated sometime S:
The way this video explained such a complex topic in such a comprehensive way is astounding. It took something that I likely would have found quite boring, and made it very intriguing. I'm excited to see more from this channel. I just found this creator recently, and so far not a single one of their videos has been uninteresting or unentertaining. Honestly surprised they don’t have more subscribers, they really deserve a larger fanbase.
Finally! A sane exploration of color theory (and more). I have been interested in this topic for a long time, but all other vies are so shallow they are entirely useless or they are completely unhinged. Please more videos on this subject!
About a year ago, when making a terraria mod, I ran into this exact issue of some colors looking brighter than they actually are on hsl/hsv. After a lot of looking through stack overflow and some other sites I ultimately gave up because I kept finding solutions that weren't what I wanted or just incomplete ones. Then you come along, make this video, solve all of the problems I had, elaborate on things I didn't understand, rebuild RGB from the ground up using the logic used for it and like idk how much else. Thank you ❤️
A tip for simple quantized grayscales is to adjust the banding to match perceived brightness (eg. 0.2-isch is mid-grey, so it should be mapped to the input value of 0.5). This helps to increase the resolution for low-mid values at the cost of mid-high. Basically each band will seem to have a more equal (linear) difference in brightness.
To add headaches to the equation: either if you're rendering in a computer or if you're illuminating something in real life, different SPDs even if perceptually identical will yield wildly different results when interacting with the materials and colors of your scene. Which also makes most LED white lights suck big time as their SPD is very poor (mainly three wavelenghts mixed together) versus incandescent light or natural light (which has the SPD of a black body)
@@exmello yeah, any iridescent object will throw off the whole model... Anything that refracts and bounces light will affect the raytraced path and give a different value for the perception model, some cases are manageable, others are not... I suggest trying to raytrace a gemstone called labradorite in different lightings and positions...
Note that with modern white LED lights, there's a specification called Color Rendering Index (CRI) that gives an approximate value for "similarity to natural light" and most LED lightbulbs have 90+ CRI these days.
@@nikolaihedler8883 CRI gets muddy as well, since it is an average of a bunch of different color indices combined (described as R values 1 through 15). If you look up the CRI data sheets of most LED lights, there are a few R values that are usually severely lacking - notably R9, which is the rendering index for red (red diodes are historically much weaker than other colors).
I LOVE the way you visualise and explain all the concepts of light, radiance and luminance in the video. Had myself lota of thoughts recently about how all those connect and how they affect our perception of colours and this video ties them all really nice together. One of the best educational videos I've seen, thank you!
For anyone curious about "blueness" being its own thing in the CIE, something I want to point out as an artist is that "darker" shades almost always have a hint of blue added when shading in art. *How* blue (or even if it is blue at all) depends on stylization, for instance cartoony styles will generally add a LOT of blue to shadows, but even artists who are trying to be photorealistic will add a touch of blue to their shadows. So as someone who has been manually messing with RGB values for digital art for almost 10 years, "blueness" as a metric came across as very intuitive to me.
So, when you talked about chimerical colors and that being a way to differentiate people's differing cone amounts, that reminded me of the concept that our color perception may not be the same and different colors may look differently to different people. (Your red is not my red) If this could be measured with tests or the like, wouldn't it be possible to adjust colors so that that fits, hypothetically aligning the colors of one person to match the color perception another person would have on the image? That could be a really cool way to share something of an individual's living experience to each other.
Howwwww have I not seen this video sooner??? This video explains so well the rabbit hole I have been diving down for the past several years, time to send it to all my friends! What is said in the video is all true, from my own research. Colorimetry is DRAMATICALLY understudied, and if bjorn ottosson can turn the entire field on its head on his own, so can you!!! Go forth and study. For me, I’m using perceptually uniform colorspaces, perceptual difference functions, and colorblindness simulation all together to generate highly accessible color palettes! I mostly need it for non-aesthetic reasons (data visualization), but with a couple tweaks it should also be useful for artistic purposes.
the shading with oklab palletes is amazing 35:00 the third one is so balanced on the darker shades and have such a nice contrast with the brighter tones, its by far the best on my opinion (factually right and morally correct)
I think the middle one looked kinda nice, but I also think if you made a palette in the third space with the same starting color, that would be nicer still.
so I made the image based on the second palette. I totally agree that 2 lacks contrast, and my usual workflow would include altering the contrast after the fact in ps, however sticking to the generated pallete was kinda the point of the challenge 3 looks great overall, but the train, which is essentially the main subject of the image, really doesn't contrast well against the background - if I'd used 3 initially, I would have definitevely picked a different color for that
1 is the best looking by far, since it has more pleasing and intense hue shifts whereas the latter two look almost sepia toned. I don't think that speaks to the merits of the respective color spaces though since the palettes were generated with the same parameters. Some amount of matching would need to be done for a proper comparison.
I think the third one is more smooth and balanced but I prefer the first two as they are much more visually interesting and have more vibrancy/hue and contrast which actually serves to benefit it in my completely objective and unbiased opinion (i do not use eyes i am an objective omniscient being)
This video is SO good! Thank you for covering color theory in a way that's neither vague nor overly complicated... I especially appreciate you mentioning so many different terms, which really helps me when searching for further information!!
There's something about the "Black Scene" bit that you do in your videos that I love. That something is as an editor I'm not a fan of having to look for a specific visual for every little thing. "Black Scene" avoids having to do that too much whle acknowleging that lack of visual. It funny to me, like dry humor. I don't know if it was intended to be humorous in any way or just to serve the simplest function.
*sitting enraptured for 37 minutes, rewatching bits, making coffee and hugging my knees trying to understand uh huh, uh huh. *opens clip studio "if you're using clip studio paint, you're out of luck." oh. *closes clip studio I'm gonna go sit outside lmao. Those alvvays shirts are super dope though!
This reminds me of back when my family got our first PC with a black and white CRT monitor. Using MS Paint, I used to do all sorts of drawings and, due to the monitor, could only pick shades of grey, not knowing what the actual colours were. Eventually when we did get a colour screen, I could see the pictures for what they really were. Absolute abominations.
man i really can't wait for a usable OKLAB color wheel for clip studio because it does articulate what i've been sensing within the process as someone who's in way too deep with the color theory bullshit
i've literally been having shower thoughts about everything in this video for months now cause I'm sick of dealing with RGB and how ugly it is (to my overly trained shader/gamedev brain) and this is just the perfect crystalisation of everything I've been thinking of (and it's super god damn vindicating knowing other people are coming to ideas and solutions that match my own thoughts, I'M NOT CRAZY)
16:55 It's not just you. A person with perfect color vision should, indeed, see a magenta circle and not a red one. This is because staring at the green circle fatiques the M (green) cones and leaves, not only the L (red) cones, but also the S (blue) cones fully functional, resulting in a mix of red and blue. I'm not sure which Wikipedia article you were looking at, but the article I found for afterimages states that green results in a magenta afterimage.
Oh my god this video was so amazing. I make generative art as a side hobby so I know a lot about shaders and color spaces and whatever, but I learned an INCREDIBLE amount from this video. Thanks!
traditional / digital Painter here; It is even more difficult when choosing colors for a complex scene. Because our color perception fools us. The way we perceive colors is affected by the temperature of the light sources, and also affected by the colors of the objects and the environment in the scene; and to make things worst our color perception is designed to recognize the local colors of objects even when the real mathematical colors don't match our perception. And it gets even worse by the fact that color temperature affects light and shadow separately (our eyes recognize them as different light sources) and our eyes can recognize even small discrepancies of light and shadow temperatures: When a painter says "these colors look muddy" what they really mean is: A light or a shadow is not the right temperature. Or even more accurate: "the parts of the object hit by the direct light source (what we perceive as light) and the diffuse light source (what we perceive as shadow) hace color temperatures that don't match their respective light sources. I make this last distinction because what most artists call "shadows" are simply a dimmer diffuse or lower light source, true shadow is just black. Everything that we can perceive as a color has a light source. and this example was an over simplification , because in real life, scenes have multiple light sources of different types and the equation of color balance can just get infinitely complex.
HWB is a completely overlooked one, that I think also fixes a lot of the problems, made by the guy who made HSV, in order to make stuff more intuitive to use, and it's fun because the mental model is just "mix in white and/or black paint"
yea in addition to OKLCH, there’s also OKHSV and OKHSL, which somewhat compromise the accuracy for better use in color pickers, like having saturation (ratio of chroma to max chroma) rather than chroma. imo it’s the best color space family in terms of like pareto optimality
I was literally doing research on that color theory stuff like 2 months ago, it's so nice to see a well done explanation of it, because oh my, am I bad at looking up good sources
wooo HSV and yeah imo OKLAB looks fantastic for modern full quality visuals. HSV has an awesome retro vibe that's perfect for pixel art (also imo). I like how this parallels other audio factors too: kinda analogous to the nonlinear behaviors behind saturation/compression/etc. Find the right nonlinear function for mapping the 0-1 interval and things fall right into place.
35:23 Well, I'm no digital artist, but I work in Computer Vision, which involves signal treatment and/or Deep Learning, and I suspect all this video could prove useful! Not to mention, I loved the video. It's a fascinating subject! In my experience, input images for Deep Learning typically use RGB or HSV/HSL values. I wonder if there's value in using the OKLAB colormodel to facilitate training in certain tasks... Something to think about, which might cost me some nights of sleep ! :D
Do share whether you explore this at some point - I've done research in ML before and am fascinated by how massive a difference just getting our units right can make.
The first result when you google Acerola is a site about the health benefits of a type of cherry by the same name as well as an ad for supplements of the cherry.
"Where do you think Color Theory lies on the complexity spectrum?" -eyes are immediately glued to the "Complex" label~ "That's riiight~" Hahahahahahaha, thank you for the vindication. xD I've studied color theory and as a kid even had a natural affinity for choosing great colors, but dang.... When you actually try to get into it and think about it.... It's been a struggle. Then again.... wow, this video might have actually helped me. Thinking back, as a kid I was using crayons, not digital colors, so my options were a lot more limited. That might be why I was able to choose much nicer colors much easier.
Dude, this video has me absolutely GLUED to the screen. Ridiculously high quality content and I'm learning a ton. Massive respect to your researching and presenting skills.
The train images are a great example of how the different options can be used to change the feeling of the picture. 1 - is very pretty and light, like a sunrise/sunset 2 - looks smoky, or foggy, or maybe it's sepia because it's an old faded photo 3 - is deep and dark and could easily be shot late at night
There used to be this podcast on Twit (This Week in Tech) Maxwells House. Every week he would mostly talk about a different area of color theory. As he said, if figured he would take a 6 week course and be all over color theory. Instead he was still studying it 30 years late and had only just scratched the surface on it. This is a field that goes DEEEEEEEP!
What I'm thinking is that posterization should be based on a good color space as well. Basically, separating color space in voxels that have the same volume in a perceptual difference metric (and making the separation lines aligned with that). One of the issues of reducing color resolution by just separating RGB space is that luminance only gradients will have artificial hue bias due to the aliasing of the gradient line in the reduced color space.
It's crazy how this super niche topic makes me stop doing whatever I was doing and just sit there with my mouth open being amazed at everything you say. Favorite video so far! Thank you!
such a fantastic video with excellent references! i particularly loved the comparison you shared at 32:32. im sure we've all seen gradients like the HSV one that just felt "off", but its hard to know what it "should" look like without having seen the Oklab one. amazing! really a terrific piece of programming and scientific communication. you should be proud!
I'm a shitty artist with a fulltime job and no quiet workspace at home. So I don't have patience for colormatching across the HSL wheel anyway. Though I have no need for photorealism, I have done some cartooning. I figured the simplest solution would be to build a spreadsheet that generates gradients based on "hues," where a hue is any color whose RGB values add up to (within a certain margin for rounding errors) 383: halfway between black and white (Something like FF4040 lolor 808080, to name 2 I can do the math for off the top of my head). Obviously. That's what everyone does ... I'm not a weirdo. It gave me a much more real blend of primary colors 'cause I have no idea what FF0000 is supposed to be, but I've seen colors like FF4040 irl. On a more practical level, the gradient generator lets me put in any RGB color, even if it's not near a "hue" tone. So it then tells me that color, white, black, 7 highlight colors, and 7 shadow colors. And that's SUPER useful to me.
staring at that green circle made me see a light pink cricle, but then a persisting gray shadow square floating in my vision for a while after the pink disappeared. am i going to die?
Ughh this is so good. Your content is like that very specific itch that feels so good to scratch. This video reminds me of that Bloom video from Ande the Great.
The first 100 people to use code ACEROLA at the link below will get 60% off of Incogni:
incogni.com/acerola ! #ad
Back to basics with this one. Thanks for 100k subs! I will be making the 100k sub special in November, please look forward to it. Check out my patreon for the potential topics for next month though!
I love how this video shows you how to Deep-Fry your Memes.
❤❤q❤
Honestly, if anything, Incogni is *more* insidious than the data brokers that it supposedly protects you from. Data brokers might profit off your personal information, but Incogni profits off your perceived anxieties over people profiting off your personal information. It's fearmongering as a business, just like VPNs. The reality is that you are not that important and that there is nobody specifically looking up your personal information to do bad things with it. Literally nobody. Stop freaking out about things that happen only in your imagination.
begging like a dog. but that "money" can't buy you a little girl bride. Which is what money was invented for in sumer. Id10r
with this pace you might want to do 100k and 200k special in November :-D congrats 🎉
Just wait until you get into color theory for print. If you think RGB sucks you should try CMYK
Fr
Exactly, like me who started decades ago and still try to make sure im understanding the concepts of this sh*t to this day and forced me to study pholosophy cause all i wanted was to actually understand it.
it just drove me insane.
mogswamp my beloved
a mc youtuber here lol
I cry, pantone laughs
For those who want to try out the color generator, it can be used in the wayback machine in the internet archives.
Tysm!!!
It's been removed from the desc so a re-share would be appreciated by future readers.
@@ultimaxkom8728 afaik sharing links in the comments is against TOS or something.
._.
what's the link?
This video feels like a severe case of ADHD that managed to condense years of Art academy in 37 minutes while still talking about the same topic, it's glorious!
One of the best parts of acerola's videos to me is that they usually progress the same way my brain works. Start talking about how to create pleasing color palates at random and after a while is discussing why posters design aren't as creative and informative as they used to be, just to stop and think "way why did i get here in the first place?! Oh yeah random colors!"
Not what ADHD is
Honestly one of the best videos on color science on youtube today (and I've watched a lot!). Subbed.
Literally Blender Guru?
no way the nft guy lol
Acerola is the only person to make me drop everything i am doing to watch a 37 minute video about computer graphics
I was supposed to be doing hw rn so real
Fr
Exactly, how does this man not have 500k 😭
Same
Well maybe sebastian lague too
13:55 : I’ve heard that mantis shrimp perception actually just uses “which of the 16 cone types is responding the most in this region”, so their color perception is, weirdly, just 16 colors, rather than a point in a 16-dimensional space.
that'd make sense! the 16 dimensional space sounds cooler though so i will simply pretend that is the case
I absolutely love when evolution does stupid shit like this. I mean, it's pointless; why not just have a bit of colour blending and get sooo much more visual acuity. but does it work, even if only just _well enough,_ at least? yes, apparently. and so it do be like that.
(for those interested in other wonderful evolutional fuckery, go look up how giraffes have a nerve that goes from the brain, down to loop behind a big artery near the heart, and _up to the mouth again_ - this is the case for all mammals, but with giraffes it's particularly egregious. they have actual mouth-movement lag)
@@mnxs I guess they might not have big enough brains for combining colour, so they got better eyes instead
God did a bit of trolling when designing these weird eyed freaks/beauties
@@mnxs yup, the biggest problem this does for giraffes is that they simply can't make noises properly. their vocal cords can never operate in sync when half of it is suffering permanent ping issues.
As somebody who has gone down the digital colour hellhole over the last few years, this video is QUITE GOOD. You've done a great job at summarizing many important concepts and this is a resource I'll absolutely be sharing with folks who I need to introduce the topic to in the future. Really well done!
I agree! as a color science hobbyist this video covers a lot of information quite accurately.
The only major mistakes that I noticed were the intuition for what X, Y, and Z change about a color and the definition of chromaticity.
TLDR:
XYZ is very similar to the RGB color model in that it is additive with the X 'primary' being reddish, the Y greenish, and the Z bluish.
Chromaticity means a specific point on the chromaticity diagram i.e. color without regard to luminance.
and here's the long version:
XYZ is very similar to an RGB color model in that it is additive with the X 'primary' being reddish, the Y greenish, and the Z bluish. The catch is that the X and Z axes are perpendicular to the luminance axis which means that they would theoretically (being impossible colors) appear black on their own. Only by adding a little bit of luminance (Y) would you be able to get an idea of what they look like. You can also think of X as leading to the right in the chromaticity diagram, Y leading up, and Z leading to the lower left corner.
Chromaticity means a specific point on the chromaticity diagram i.e. color without regard to luminance. Another way of thinking of chromaticity is to remember that the chromaticity diagram is a projection of a slice of the XYZ space. Imagine a ray through the origin of the XYZ or RGB color spaces. Technically the origin is black and is equivalent to any chromaticity at 0 luminance, but any other points on a ray will have the same chromaticity while differing only in luminance.
Also went down this rabbit hole a few years ago. I ended up on CIECAM02 which since OKLAB wasn’t really that popular back then.
I wish this video existed back then because it would’ve saved me weeks of research.
In terms of doing math with colors (such as photoshop layers blending modes), do you believe it is better to work with a perceptual models such as Oklab or instead work with ACEScg or other industry-standard linear RGB spaces?
I'm new to all this theory and I am interested in any insight about the differences between these
I can only agree with Hank. I still have to find a good tristimulus dataset to convert any pure wavelength to a perceptual color, in a way that when you draw a spectrum, it's not plagues with visible bands. Meanwhile I devised my own color model, HCI (Hue Chroma Intensity) that arguably works better than *Lab color models, but is computationally very expensive to invert (requires binary search).
@@Uncl3M3at Both have their value. It's about using the right tool for the job. At the end of the day, your output will be a formed RGB image, the steps to get there are of course entirely dependant on your specific situation.
It's preferential to use OKLab to calculate gradients for the reasons explained in the video, but you're going to have to convert those generated values to RGB values so you can actually use them in a bitmap image or display them on a screen.
As for the blend modes, those also have value as they are used for artistic effect and produce a (presumably) desired response. They're equally valid.
i suspect one major reason HSL palettes were still frequently picked is that it will sometimes "incorrectly" give you colors with more chromaticity. for a nice gradient, maybe you want one of those three parameters to remain constant, but for an artistic palette i would want at least one color that pops over the rest. if the OKLab generator was to occasionally provide a nice vibrant color along with the rest of the gradient, you might see even more people favoring it.
I second this.
as another digital artist, this is an amazing insight! i was a little thrown off by how much i liked some of the hsl palettes, but this explains it as well as how much more consistent oklab was in comparison!
Honestly, if more math teachers teached using graphical examples of what you can do with it like you do, I think quite a bunch of people would be more interested in mathematics, amazing video as aways Acerola.
Honestly I still prefer the topics being introduced in classes the same way. I think the graphic examples and uses are more helpful after to help solidify the concepts. The problem with that is educators have basically no time to do both, leading to an unsatisfactory experience regardless.
If only more math teachers spent hundreds of hours preparing every lesson… making videos like this is hard, long, tedious work.
Honestly, having each class of students being taught by individual teachers is inefficient. The main source of knowledge should be the best of prerecorded lessons and an on-site teacher is just there to help with questions that can't be googled quickly and to keep distractions down.
Most digital artists who have a more painterly art style usually start with a grayscale base just to establish the contrasting values, then we add color on top of that using adjustment layers that doesn't affect the values, only the hues. That way we keep the dynamic contrast and can still use whatever color sliders we have at hand. Coloring a grayscale base isnt a universal thing ofc, but it has it's roots from oil painting techniques where artists paint a sepia toned underpainting that serves the same purpose.
Interesting! I would imagine that using a perceptual color space to calculate the adjustment layers would really make a difference there. Changing the hue and saturation while keeping the HSL/HSV "luminance" or "value" number the same is going to change the perceived value, in some cases quite a bit.
@@BrooksMoses I was thinking the same thing! Definitely going to experiment with that colour space since I use Krita.
That is so wrong, i don't even know where to start
It truly amazes me how you can take such a seemingly small problem and dig into it to explain everything that is actually involved in solving it, which is a lot, while still managing to keep it all connected and fascinating. As a computer science student who is also into graphic design and painting, you genuinely make me consider taking a professional step towards graphics programming, which is something I didn’t even know about before I stumbled upon your channel. Thank for the inspirational videos and keep doing it like you do !
Same. You said it all :D
My senior capstone project in college was about palette selection from still images. I don't know anything about shaders or color theory, but because I've your videos I have been inspired to dust it off and see if I can get it to compile again.
Edit: after jumping through a lot of hoops to make JavaFX work again, I got it to compile. 6 years ago me was not a good programmer. The basic concept was to analyze an image and select a limited color pallet that would keep that average (mean) color of the entire image the same. This mostly produced extremely ugly images, but occasionally made some really neat effects.
im a biology student, but the way your videos grab my attention is absurd please never stop
When Acerola starts talkin cell biology:
Neuron activation
so I'm also a bio student, and I seek info. The RGB color model means that the screens don't show all colors we can see, or are those colors outside it the so called imaginary colors
@@HoneyDoll894 Both, if you're talking about the sRGB triangle. There are colors outside of sRGB that we see normally and sRGB can't replicate them due to the LCD screen limitations. You could still use those colors if you were to paint them in real life. There are also imaginary colors that only our eyes can see and we couldn't ever replicate a paint for those colors because those imaginary colors only exist in our minds under certain contexts like cone fatiguing.
@@HoneyDoll894 is that a Phosphophyllite profile picture? that's awesome
@@HoneyDoll894 It really doesn't matter much what the visible spectrum for us humans is. The physical ability of LEDs or CRTs has its own intensity graph, all of them nonlinear. This is adjusted by filters on the electronics level to map that to a more usable spectrum. More translation is applied (try to switch the profiles of your monitor). Then we have screen-space colors and all the color models you mentioned. Another level of filters, in your drawing/video program will help you to produce/adjust content for print media, for web or other forms of distribution, like movie theaters ... you get what I mean, its simply overwhelming:)
Direct RGB to, lets call it, human eye-receptor-mapping would be non-linear, because of our evolutionary higher sensitivity shaped by evolution, e.g. the greatest light sensitivity for spectral colors in the yellow-green range (hey? when did you eat a nice meal of grass, the last times? hehehe). The same applies to an arbitrary, for example 0-255, 32bit integer value encoding colors and applying them as a voltage/phase-angle-modulation/etc to light emitting technology. Those aren't linear either. Any of those filters in this possible long and complicated chain (from your Image data through the monitor into your eyes) help to make the experience more natural and also standardized. Well, as I said, there are several techniques for different applications. Viewing is a whole different approach than printing media.
Btw that all of that does make really sense, one has to CALIBRATE his monitor. Mostly no one (artists and professionals of course excluded here), no normal computer user has such a device or cares about that. Color receptance is highly individual and can even change with your personal mood or, your hormones or the time of the day (most people don't have a standardized studio light background either ... but our lovely sun:) ). So this can be a topic of fights to the blood (of some nerds?) and also the most inconsistent science of all (depends of course if you "eliminate" those pesky humans out of the equation, hehehe).
I wish you both a wonderful time and big discoveries and exciting insights during your studies!:) (This text was written in a dark grey color. Hmm, I should clean my monitor ...)
It is just me, that thing we enter the wrong class
I thought it was going to be about color theory
@@sunshines4898
No way, me too lol
I've always been struggling with colors because I am colorblind. So learning the theory has always let me understand everything more clearly.
now you can see more colors right
Same, I didn't know how bad until I had a class where we were supposed to create different color relationships with color swatches and I had such a hard time with it.
i don't want to whine too much about one of the admittedly less debilitating disabilities one can have but i really do think people underestimate just how much colorblindness impacts the way you process, and in turn depict, the world.
like holy shit guys did you know that Light? i sure fucking didn't.
@@northropi2027 No see, as an artist who can see in all color, I swear to God I had a dream once where I went colorblind and was absolutely devastated.
Here's an abridged version (it's really long I'm sorry ;-;) :
9/17/23 Sun.
This was mostly in the first person. I woke up from a coma to my mom feeding me cornmeal mush or something. Everything was a warmish blue, and was dark. Looking in the bathroom mirror, I saw that my left eye was missing. Everything remained blue as I went to ask mom what happened, and she reluctantly told me that I'd somehow cut it off before passing out and falling into the coma; the surgeons couldn't help much.
With only one eye, I lost my depth perception, which was to be expected, though it was still upsetting. Everything then turned black and white and I thought (for some reason) that I needed both eyes to percieve color; this only further devastated me.
I had an ipad or tablet on hand, and opened up to a color wheel. I tried imagining the colors overlaid on top of it, but it felt horrible. I searched up art tutorials which didn't require color as much (charcoal, graphite, micron pen, anatomy, shape, texture, etc), "but with the media so rich in color, every other tutorial regarding gorgeous sunsets, vibrant floral tones, and other color palettes were hard to sit through, and I skipped through all of them."
The rest of the dream was kind of weird because one of my legs was replaced with a lion leg and there were some interesting shenanigans, a water park, and two siblings having a rap battle...? Then there was one section that went, "The walking was inconvenient for sure, *but not so bad as the vision loss.* I would see flashes of color, though now I'm not sure if I was just imagining it. When I would think of things, I'd think of them in color, but I was already starting to forget exactly what they looked like--which particular hue it was, how warm my mom's face was, which gradient went over the shadows," this was followed by a look in the mirror again revealing that a wrinkly little eye was growing back somehow.
Inaccurate as the actual aspects of colorblindness were, I think it's safe to say that this dream was within the top 5 tiers of my list of nightmares. It was awful.
I've been doing digital art in photoshop for a while now, and I always KNEW that the included color picker was frustrating because of the inherent values of some hues (e.g. fully saturated+luminant yellow is still a light value than fully saturated+luminant blue). I knew what I wanted in an updated color picker but couldn't even think how to describe it, and OKLAB is super interesting. This video had me glued to my screen and was hilarious the entire time, and I feel my brain growing wrinkles. This was super interesting, thank you for this video.
You're such an excellent communicator! It's cool that we're coming at the same problems from different sides: my background is in art and I've been approaching computer graphics understanding perceptual color first, and now trying to learn how people have tried to map it to the tech we use to view it.
As soon as I realized we were about to dive into color spaces and color theory, I buckled my seatbelt and called my parents in case I didn't make it. Glad to see you didn't self destruct half way through, only your clone :)
1. I saw a pale magenta (like #ffa1e4)
2. As a digital artist who learned color/drawing almost entirely digitally, this is a really cool video to watch! Digitally picked colors are my favorite part of the process, and while I manage to be good with digital color it is at the cost of being able to do literally anything else intuitively. And yes, HSL is useful but it can be a nightmare in terms of picking colors that we actually perceive as brighter. For me I use a special color profile to check, alongside the old fashioned eye squint to determine value. Also, I'm out of luck.
Yes I also see magenta. And it makes sense as it's the opposite of green on the colowheel.
16:40 Same here, but I'd say even lighter, like #ffd0f1
Mine was closer to ffc2fb, but it's so hard to try and find a matching 'real' color...
i definitely saw a bright saturated magenta too, but i really enjoyed the massive nothing square in the middle of my vision from the white square around the circle. the square persisted for much longer
Was hoping to find some discussion on the cone fatigue test. Think I saw a fairly saturated magenta but seemed light due to the white background.
Love the fact that the World of horror soundtrack started playing right when it got to the random palette part
"Oh, this sounds like it'll be interesting"
[15 minutes later]
"Ah yes, imaginary colors."
I felt so heard when you listed Color Theory as most complex. I tinker and obsessively toy with RGB hex values for a living. ❤ Massively useful banger vid as always.
At 16:41 I actually saw a very very light blue/cyan colour. But I also have a color blindness of two types with mid/low intensity (prota+deutera-nomalies) so that's what might cause such a difference from the expected outcome. I also really loved your video about mentioned topic. Keep the good work!
Definitely a colourblindness thing, I saw light blue/cyan too and have a mild red/green colourblindness.
Photoshop is not the only alternative nowadays. A painting software like Krita do offers Lab as well. Krita also allows to have a grayscale luminance view aside the canvas allowing for direct corrections.
yeah i forgot krita exists, aseprite also has options for lab i think
Aseprite gang
oh crap that's what I use! So you're saying I can select ...Lab? as the colorspace in Krita?
@@CrescentUmbreon yes L*a*b* is available at least since 4.0 so it's been a while. Coupled with the ability to display different views (like grayscale or mirrored) this is really easy to directly paint without a complex preliminary sketch.
Though I don't know if there is some view to get the overall median color/luminance of an image, to check its balance.
@@PainterVierax Thank you, I'll have to research it!
That oklab gradient at 32:35 is sooo satisfying, it absolutely blows HSL hilariously uneven luminance out of the water what the heck I need to use this thing RIGHT NOW.
As a student in computer graphics and image analysis, your videos are always entertaining and informative but this one especially so; OKLAB is a game-changer.
Babe wake up acerola released a new video
THANK YOU! Finding a way to map Human Colour Perception to a screen's representation of it has been so utterly difficult. Not just that, even understanding that issue is so tough and only now after watching this video do I feel like I understand where my biggest issues with understanding colour have come from. This will help me so so much
I love color theory. There are so many strange things that happen. Like if you mix the primaries RGB, which are additive colors, with the secondaries CMY, which are subtractive, you'll find that they oscillate between a color that was 50% lighter than you expected, neutral, and 50% darker than expected. When you mix green and red light, you get yellow, because the two colors add light together. What do you get when you mix green and red pigment/paint though? Most people would say you get black, but you don't, you get a brown. But why should the color, the hue be any different from when we were mixing light? It shouldn't change the hue, it should just make it seem darker. That's when it hit me, I had just watched a video on how brown isn't a real color, it's just context. Any reddish orange to yellow color when darkened can be considered brown. So when we mix our red and green light we get yellow, and when we mix the red and green pigment, they are subtractive and darken the yellow! THAT'S why it looks brown! You DO get the same color when mixing pigments as you do mixing lights. It's just darker. If you mix two subtractive colors (ie: red and green), it will be darker, mix an additive with a subtractive (ie: yellow and red) it will expected brightness, mix additive with additive (ie: yellow and magenta) it will be brighter than expected. And green is HUGE. It's is so massive compared to the other colors. You try to mix a color halfway between green and yellow, and you get green lol. You try to mix a color halfway between green and blue and you get green. It's crazy how much further toward the adjacent colors you need to go from green to get those tertiary colors.
Combining pigments is not the same kind of colour theory you have to consider what is taking place there is chemistry, it depends entirely on how those things react as to what the result will be not simply what their apparent colour is.
well these were both super interesting comments 😱
@@iisthphirchemistry? no. you're implying that there is a chemical reaction happening between the components of the paints. that is not the case. it's just pigments getting _mixed,_ that's all.
It depends on how the pigments react, that would be a chemical reaction. It could be extreme or basically nothing it just depends on what is in them. These days I think most types of paints are made such that they react very little one of the reasons being so you can mix them with predictable results. There are others that are made to react with the surface they are painted on for instance. Historically though pigments were made from all sorts of strange things many of which were not as stable and so did react to some degree, with each other or simply by changing over time from oxidation. Sometimes it may be necessary to paint in one colour to have it end up being a different one. As, I understand, was the case with some of those used by the Italian's during the renaissance.
@@iisthphir it seems unwise to make a historical parallel to modern paints; the latter are often polymer-based or otherwise examples of modern chemical technique and considerations.
if two paints are mixed, they are usually from the same manufacturer and of the same type, and thus have the same base chemical composition, _except_ the pigments - thus compatible. but sure, if you mixed a two paints of radically different compositions, weird stuff might happen.
I still doubt it would attack the pigments themselves much, though. given modern environmental, toxicological and general safety requirements, I believe pigments wouldn't be particularly reactive anyway - not to mention that it's just not a very desirable trait for a supposedly stable pigment to have.
also, the chemical reactions that might happen with the painted surface are one thing, and shouldn't affect the pigments by design. if the paints share a common base composition, and that base won't attack one pigment when it's by itself, why would it attack when it's mixed with something with a different pigment? besides, any non-trivial colour paint would most likely be a mix of any number of pigments anyway.
12:22 Fun fact: because of red’s lack of luminance red light does not cause your vision to switch to “cone mode”, meaning that when you turn off a red light your eyes with still be adjusted to the dark! This can be very useful at night.
I was so intruiged what the mystery behind those three pallette generators was! Thank you and I'm now going to have "OKLAB" in a drawer somewhere in my mind in case I ever need it!
This was a psychological horror movie. Anyway, amazing work, I can’t imagine the amount of research that goes into these videos, as a digital artist this was fascinating, and gave a nice explanation of the problems with HSL that I think we all mostly assumed were just something we’d have to live with, and also why they exist
A big problem I come across is how all my screens are not calibrated by default, and I assume most viewers viewing my art do not have professionally calibrated screens either. So when creating something, I usually tend to view it on different screens to make sure it looks kind of okay on all of them, but it's impossible to be too picky about this stuff. Maybe I should actually get my screen calibrated sometime S:
The way this video explained such a complex topic in such a comprehensive way is astounding. It took something that I likely would have found quite boring, and made it very intriguing. I'm excited to see more from this channel. I just found this creator recently, and so far not a single one of their videos has been uninteresting or unentertaining. Honestly surprised they don’t have more subscribers, they really deserve a larger fanbase.
Finally! A sane exploration of color theory (and more). I have been interested in this topic for a long time, but all other vies are so shallow they are entirely useless or they are completely unhinged.
Please more videos on this subject!
25:30 made me laugh SO hard. Poor Acerola...
About a year ago, when making a terraria mod, I ran into this exact issue of some colors looking brighter than they actually are on hsl/hsv. After a lot of looking through stack overflow and some other sites I ultimately gave up because I kept finding solutions that weren't what I wanted or just incomplete ones.
Then you come along, make this video, solve all of the problems I had, elaborate on things I didn't understand, rebuild RGB from the ground up using the logic used for it and like idk how much else.
Thank you ❤️
A tip for simple quantized grayscales is to adjust the banding to match perceived brightness (eg. 0.2-isch is mid-grey, so it should be mapped to the input value of 0.5). This helps to increase the resolution for low-mid values at the cost of mid-high. Basically each band will seem to have a more equal (linear) difference in brightness.
To add headaches to the equation: either if you're rendering in a computer or if you're illuminating something in real life, different SPDs even if perceptually identical will yield wildly different results when interacting with the materials and colors of your scene. Which also makes most LED white lights suck big time as their SPD is very poor (mainly three wavelenghts mixed together) versus incandescent light or natural light (which has the SPD of a black body)
yeah this is why rgb ray tracing is so wildly diff (and wrong) from spectral ray tracing in some specific contexts
Can you give an example? Is there a common material that's wildly off?
@@exmello yeah, any iridescent object will throw off the whole model... Anything that refracts and bounces light will affect the raytraced path and give a different value for the perception model, some cases are manageable, others are not... I suggest trying to raytrace a gemstone called labradorite in different lightings and positions...
Note that with modern white LED lights, there's a specification called Color Rendering Index (CRI) that gives an approximate value for "similarity to natural light" and most LED lightbulbs have 90+ CRI these days.
@@nikolaihedler8883 CRI gets muddy as well, since it is an average of a bunch of different color indices combined (described as R values 1 through 15). If you look up the CRI data sheets of most LED lights, there are a few R values that are usually severely lacking - notably R9, which is the rendering index for red (red diodes are historically much weaker than other colors).
I LOVE the way you visualise and explain all the concepts of light, radiance and luminance in the video. Had myself lota of thoughts recently about how all those connect and how they affect our perception of colours and this video ties them all really nice together. One of the best educational videos I've seen, thank you!
For anyone curious about "blueness" being its own thing in the CIE, something I want to point out as an artist is that "darker" shades almost always have a hint of blue added when shading in art. *How* blue (or even if it is blue at all) depends on stylization, for instance cartoony styles will generally add a LOT of blue to shadows, but even artists who are trying to be photorealistic will add a touch of blue to their shadows. So as someone who has been manually messing with RGB values for digital art for almost 10 years, "blueness" as a metric came across as very intuitive to me.
So, when you talked about chimerical colors and that being a way to differentiate people's differing cone amounts, that reminded me of the concept that our color perception may not be the same and different colors may look differently to different people. (Your red is not my red)
If this could be measured with tests or the like, wouldn't it be possible to adjust colors so that that fits, hypothetically aligning the colors of one person to match the color perception another person would have on the image?
That could be a really cool way to share something of an individual's living experience to each other.
Its always cool for you to just casually explain collage level math like nothing.
Howwwww have I not seen this video sooner??? This video explains so well the rabbit hole I have been diving down for the past several years, time to send it to all my friends!
What is said in the video is all true, from my own research. Colorimetry is DRAMATICALLY understudied, and if bjorn ottosson can turn the entire field on its head on his own, so can you!!! Go forth and study.
For me, I’m using perceptually uniform colorspaces, perceptual difference functions, and colorblindness simulation all together to generate highly accessible color palettes! I mostly need it for non-aesthetic reasons (data visualization), but with a couple tweaks it should also be useful for artistic purposes.
the shading with oklab palletes is amazing
35:00 the third one is so balanced on the darker shades and have such a nice contrast with the brighter tones, its by far the best on my opinion (factually right and morally correct)
I think the middle one looked kinda nice, but I also think if you made a palette in the third space with the same starting color, that would be nicer still.
so I made the image based on the second palette. I totally agree that 2 lacks contrast, and my usual workflow would include altering the contrast after the fact in ps, however sticking to the generated pallete was kinda the point of the challenge
3 looks great overall, but the train, which is essentially the main subject of the image, really doesn't contrast well against the background - if I'd used 3 initially, I would have definitevely picked a different color for that
third one is way too dark in HDR, can barely distinguish the dark colors
1 is the best looking by far, since it has more pleasing and intense hue shifts whereas the latter two look almost sepia toned. I don't think that speaks to the merits of the respective color spaces though since the palettes were generated with the same parameters. Some amount of matching would need to be done for a proper comparison.
I think the third one is more smooth and balanced but I prefer the first two as they are much more visually interesting and have more vibrancy/hue and contrast which actually serves to benefit it in my completely objective and unbiased opinion (i do not use eyes i am an objective omniscient being)
This video is SO good! Thank you for covering color theory in a way that's neither vague nor overly complicated... I especially appreciate you mentioning so many different terms, which really helps me when searching for further information!!
I always learn so much from your videos! Had no idea about this, but now I understand why I always have so much trouble picking colors using HSV
was just here to get better at art and got a whole physics lesson waow. also love the use of persona music, great video
There's something about the "Black Scene" bit that you do in your videos that I love.
That something is as an editor I'm not a fan of having to look for a specific visual for every little thing.
"Black Scene" avoids having to do that too much whle acknowleging that lack of visual.
It funny to me, like dry humor. I don't know if it was intended to be humorous in any way or just to serve the simplest function.
lol sometimes it's used for comedy other times to indicate something complicated or unpleasant or just i was lazy and needed to fill space
*sitting enraptured for 37 minutes, rewatching bits, making coffee and hugging my knees trying to understand
uh huh, uh huh. *opens clip studio
"if you're using clip studio paint, you're out of luck."
oh. *closes clip studio
I'm gonna go sit outside lmao.
Those alvvays shirts are super dope though!
Watching this in class with no audio (and no sub). what a fever dream. Amazing editing
Acerola is the most attractive and studious man I know. He is so smart and funny, I long to be just like him one day.
34:15 suprise scug
Also very concise video, it all went over my head by it was still fun to watch
This reminds me of back when my family got our first PC with a black and white CRT monitor. Using MS Paint, I used to do all sorts of drawings and, due to the monitor, could only pick shades of grey, not knowing what the actual colours were. Eventually when we did get a colour screen, I could see the pictures for what they really were. Absolute abominations.
Please PLEASE never delete your colour randomizer site. I always have problem with colours in my art, so this is pure gold fo me. Thank you 🙏
man i really can't wait for a usable OKLAB color wheel for clip studio because it does articulate what i've been sensing within the process as someone who's in way too deep with the color theory bullshit
You just activated an old memory of Maria Shugrina's papers on palette mixing and palette replacement. Thanks !
i've literally been having shower thoughts about everything in this video for months now cause I'm sick of dealing with RGB and how ugly it is (to my overly trained shader/gamedev brain) and this is just the perfect crystalisation of everything I've been thinking of (and it's super god damn vindicating knowing other people are coming to ideas and solutions that match my own thoughts, I'M NOT CRAZY)
How exactly is the RGB model ugly? You mean naively generated palettes in the sRGB space as discussed in this video tend to be ugly?
I love broad overviews of dense topics. When I see related information again, I’ll have context to more easily understand it
Quite excited for the Ethics in graphics programming video
16:55 It's not just you. A person with perfect color vision should, indeed, see a magenta circle and not a red one. This is because staring at the green circle fatiques the M (green) cones and leaves, not only the L (red) cones, but also the S (blue) cones fully functional, resulting in a mix of red and blue. I'm not sure which Wikipedia article you were looking at, but the article I found for afterimages states that green results in a magenta afterimage.
Oh my god this video was so amazing. I make generative art as a side hobby so I know a lot about shaders and color spaces and whatever, but I learned an INCREDIBLE amount from this video. Thanks!
34:18 Panko my beloved
Awesome video Acerola, this has enlightened my mind and opened my eyes.
Awe. Palette generator link doesn't work.
Can anyone re-share it? It's removed from the desc but we should still be able to use it by the wayback machine.
RAINWORLD SLUGCAT at 34:15
"imagine there are different colors"
i felt that
traditional / digital Painter here; It is even more difficult when choosing colors for a complex scene. Because our color perception fools us. The way we perceive colors is affected by the temperature of the light sources, and also affected by the colors of the objects and the environment in the scene; and to make things worst our color perception is designed to recognize the local colors of objects even when the real mathematical colors don't match our perception. And it gets even worse by the fact that color temperature affects light and shadow separately (our eyes recognize them as different light sources) and our eyes can recognize even small discrepancies of light and shadow temperatures:
When a painter says "these colors look muddy" what they really mean is: A light or a shadow is not the right temperature.
Or even more accurate: "the parts of the object hit by the direct light source (what we perceive as light) and the diffuse light source (what we perceive as shadow) hace color temperatures that don't match their respective light sources. I make this last distinction because what most artists call "shadows" are simply a dimmer diffuse or lower light source, true shadow is just black. Everything that we can perceive as a color has a light source. and this example was an over simplification , because in real life, scenes have multiple light sources of different types and the equation of color balance can just get infinitely complex.
HWB is a completely overlooked one, that I think also fixes a lot of the problems, made by the guy who made HSV, in order to make stuff more intuitive to use, and it's fun because the mental model is just "mix in white and/or black paint"
i'm a big fan of HCL since it tracks pretty well with intuitive human perceptions of color
@@unslept_em OKLAB color space can be expressed in LCh numbers as well, it was even added to CSS along with OKLAB as OKLCH.
yea in addition to OKLCH, there’s also OKHSV and OKHSL, which somewhat compromise the accuracy for better use in color pickers, like having saturation (ratio of chroma to max chroma) rather than chroma. imo it’s the best color space family in terms of like pareto optimality
I was literally doing research on that color theory stuff like 2 months ago, it's so nice to see a well done explanation of it, because oh my, am I bad at looking up good sources
ready for 40 minutes of peak 🙏
Oh my god, I just came here because I wanted to learn how to choose better colors for my outfits in video games. I wasn't expecting color science.
wooo HSV
and yeah imo OKLAB looks fantastic for modern full quality visuals. HSV has an awesome retro vibe that's perfect for pixel art (also imo).
I like how this parallels other audio factors too: kinda analogous to the nonlinear behaviors behind saturation/compression/etc. Find the right nonlinear function for mapping the 0-1 interval and things fall right into place.
32:32 the gradient at the top is fricken supernatural, I've never seen anything like it with my eyes before. So amazingly smooth and even
35:23 Well, I'm no digital artist, but I work in Computer Vision, which involves signal treatment and/or Deep Learning, and I suspect all this video could prove useful! Not to mention, I loved the video. It's a fascinating subject!
In my experience, input images for Deep Learning typically use RGB or HSV/HSL values. I wonder if there's value in using the OKLAB colormodel to facilitate training in certain tasks... Something to think about, which might cost me some nights of sleep ! :D
Do share whether you explore this at some point - I've done research in ML before and am fascinated by how massive a difference just getting our units right can make.
13:14 "The wave enters your balls." Doesn't that sound wrong?
The first result when you google Acerola is a site about the health benefits of a type of cherry by the same name as well as an ad for supplements of the cherry.
i'm playing from behind in the search engine optimization game
Truly amazed at how much dense content you can talk about coherently over an almost 40 min vid!
It seems the OKLAB palette generator is no longer up. Does anyone know of another OKLAB palette generator.
Bump cause I wanna know too. I can only find a monochromatic one.
"Where do you think Color Theory lies on the complexity spectrum?"
-eyes are immediately glued to the "Complex" label~
"That's riiight~"
Hahahahahahaha, thank you for the vindication. xD I've studied color theory and as a kid even had a natural affinity for choosing great colors, but dang.... When you actually try to get into it and think about it.... It's been a struggle. Then again.... wow, this video might have actually helped me. Thinking back, as a kid I was using crayons, not digital colors, so my options were a lot more limited. That might be why I was able to choose much nicer colors much easier.
Sine waves in the West were originally discovered by literally recording ocean waves.
32:30 this is god tier. As a digital artist for 30 odd years, this uniformity of intensity is just pure joy
Dude, this video has me absolutely GLUED to the screen. Ridiculously high quality content and I'm learning a ton. Massive respect to your researching and presenting skills.
I love how in depth these videos are. Incredible work!
34:05 LAIN LAIN PFP ITS LAIN ITS LAIN IM GOING INSANE
The train images are a great example of how the different options can be used to change the feeling of the picture.
1 - is very pretty and light, like a sunrise/sunset
2 - looks smoky, or foggy, or maybe it's sepia because it's an old faded photo
3 - is deep and dark and could easily be shot late at night
I litteraly crave Acerola videos in my heart every single day
25 minutes in and i completely forgot that this was about final fantasy gradents lol
There used to be this podcast on Twit (This Week in Tech) Maxwells House. Every week he would mostly talk about a different area of color theory. As he said, if figured he would take a 6 week course and be all over color theory. Instead he was still studying it 30 years late and had only just scratched the surface on it. This is a field that goes DEEEEEEEP!
As someone with a special interest in Color theory in both art and physics, I will add this to my knowledge collection
This blew my mind as both a software engineer and an artist. Also the delivery format was immaculate * chef’s kiss * you definitely earn my sub
What I'm thinking is that posterization should be based on a good color space as well. Basically, separating color space in voxels that have the same volume in a perceptual difference metric (and making the separation lines aligned with that). One of the issues of reducing color resolution by just separating RGB space is that luminance only gradients will have artificial hue bias due to the aliasing of the gradient line in the reduced color space.
It's crazy how this super niche topic makes me stop doing whatever I was doing and just sit there with my mouth open being amazed at everything you say. Favorite video so far! Thank you!
such a fantastic video with excellent references!
i particularly loved the comparison you shared at 32:32. im sure we've all seen gradients like the HSV one that just felt "off", but its hard to know what it "should" look like without having seen the Oklab one. amazing!
really a terrific piece of programming and scientific communication. you should be proud!
I'm a shitty artist with a fulltime job and no quiet workspace at home. So I don't have patience for colormatching across the HSL wheel anyway. Though I have no need for photorealism, I have done some cartooning.
I figured the simplest solution would be to build a spreadsheet that generates gradients based on "hues," where a hue is any color whose RGB values add up to (within a certain margin for rounding errors) 383: halfway between black and white (Something like FF4040 lolor 808080, to name 2 I can do the math for off the top of my head). Obviously. That's what everyone does ... I'm not a weirdo. It gave me a much more real blend of primary colors 'cause I have no idea what FF0000 is supposed to be, but I've seen colors like FF4040 irl.
On a more practical level, the gradient generator lets me put in any RGB color, even if it's not near a "hue" tone. So it then tells me that color, white, black, 7 highlight colors, and 7 shadow colors. And that's SUPER useful to me.
25 minutes of color theory and then I remember we were working on colorizing FF screenshots.
staring at that green circle made me see a light pink cricle, but then a persisting gray shadow square floating in my vision for a while after the pink disappeared. am i going to die?
Ughh this is so good. Your content is like that very specific itch that feels so good to scratch. This video reminds me of that Bloom video from Ande the Great.
@Acerola your epic color palette website is not available anymore :(
Can anyone re-share it? It's removed from the desc but we should still be able to use it by the wayback machine.
My colors don’t suck, they convey my emotion and message of “I suck and don’t know what I’m doing”