Choosing colors and pallettes

[Faktablad]

I've been looking at a lot of demoscene pics lately and, though I see many things that those pieces have that mine don't, a main aspect that confuses me is color selection.  I can't seem to get vibrant, living colors in my art, so I'd like to know if any of you use certain methods to select colors.  How should I go about it?  It'd be great to hear from a lot of people, especially artists like ptoing and Helm who's color selections I've admired for a long time.

[Dhaos]

This is somewhat a complicated concept... on the most basic level 'vibrant' usually contrasting colors. During the creation of a piece of pixelart, it's palette will make continual changes as new tones are added in. You try to balance whatever it is that you are making *as* you make it. Its very hard to create all the exact tones you need before you begin a piece.

[BAISC PIXEL COLOR THEORY]
-pixels are much more exact than natural media so you must 'learn' what colors are light and dark
*the follow values are the HUE of a color (using RGB to select colors is alot more complicated... I only know like two artists who do it)
*I do not include most of the inbetween colors (teal violet, magenta etc...its not necessary right now, I'm only included commonly used colors)

[MIDTONES] (the 'base' tone if you will)
 +red:5-25hue or 340-355hue
 +orange/brown:25-40hue (the less saturation this hue has, the more brown it becomes)
 +yellow: 40-50hue
 +green: 65-80hue (darkers greens 'hue:120ish' generally look VERY bad unless they are confined to a small area on a pic)
 +blue: 200-215hue
 +purple: 220-245hue

[SHADOWS] (where light does not hit...)
 +red: 0-10hue or 330-340hue
 +orange/brown: 10-25hue
 +yellow: 20-30hue (note the shadows of yellow tend to be orange or brown)
 +green: 80-95 or 120~145 (again...there usually isn't alot of dark green in an image, even a forest scene, so this hue:120+ is ok)
 +blue: 220-245hue (note...the shadows of blue tend to be purple... perhaps its a pattern...)
 +purple: 225-245hue (purple generally is its own shadow)

[HIGHLIGHTS] (where light directly strikes an object...used to make specfic objects POP)
-This section is complicated. It involves hue shifting to figure out
-highlights are generally done by shifting your midtone hue to the next brightest color
****show hue-slider****

-red/orange/brown: maximum hue: 35-50 (aka yellow~yellowish-orange)
-yellow: maximum hue: 55-60 (any higher and you get green)
-green: maximum hue: 55-65 (yellow is brighter...thus it highlights green)
-blue: maximum hue: 160-195 (notice the huge-ass range... blue is just weird that way...most greens/teals/light blues can highlight it beautifully)
-purple: maximum hue 200-210 (aka blue) OR 300~345 (again huge range, magenta/violet can highlight purples as well)

[COLOR COMPARISON]
-yellow->green->orange/brown->red->blue->purple
-some colors are brighter than others...its just how colors are
-yellows are generally the brightest 'colors' aside from white(s)...
-colors blend best if 1)they follow the hue rainbow-thing 2)the tones are seperated by an area of shadow

Here's an example of how I go about creating a simple grass dirt and rock landscape.
[STEP 1]

-choose basic tones to work by choosing one midtone and one shadow tone for each color you want
(since I am doing grass I need two tones of green, rock: thus two tones of grey, dirt: two tones of brown)
-decide on the style/mood of the piece
(I am doing two varients, brighter and cheery, then another that is highly contrasted)
-when creating a scene you must decide what color the light is
(I use yelowish highlights and purplish shadows)
(also the highlights and shadows do NOT have to match...it creates that vibrant look you see in alot of demo-scene art)
-now lay down the basic shapes...

[STEP 2]

-again add one more tone per object (new tone somewhere between your two initally tones), detail the scene further...
-I am leaving some areas 'un-finished' becase if its generally faster to choose 6 tones per color initially (but that can be overwhelming for this explaination)

[STEP 3]

-continue adding tones till you have detailed the scene the way you want it
-focus on a few areas till you know how many tones you will need to detail everything properly
-as you add new tones you will need to re-contrast your old tones, this is where the initial color theory, at the top, comes in...
-you will have to eye-ball your scene to decide how much contrast you want, for this first mockup, I want very little, so I keep the saturation and luminance between each color fairly low (about a 5-6 point different in luminance for each tone)
-remember, you DON'T arbitraily shift each new tone down by 5 lum/hue, the difference is based on how much you want a spefic object to stand out
*ex: dirt...dirt is not highly contrasted...its flat and dirty so I keep a low lum-decrease but a keep a steady hue-shift
-remember how your lighting the scene, in this case things closer to the ground are darker and less detailed/contrasted than the things higher up, its tricky and you only learn how to do this properly by heavy study of pixel art
*here I noticed my grass was getting overly bright...blinding really, so I 'muted' it by darkening it and lowering its contrast


[STEP 4]

-this part is fun, you now must unify your pallete, even if your grass is 'green' your dirt 'brown' and your rocks 'grey' your light-source will blend them together (and the fact the scene is so #($*#'ing small)
*ex: take a high-res cg and zoom out... you'll see how everything begins to merge, blend, and distort
-pixel art thrives on controlling the color of each individual pixel, it allows you to create very clear and concise images at any resolution
*unlike a CG, where the image is generally very large capturing the entire scene, pixel art is relatively small, focusing on various elements within the piece (sprites for example)
*there are of course numerous exceptions...however it remains true that high-quality pixel art tends to have 'prettier' colors than a cg, because the pixelartist has more control over how the colors are manipulated
-various ways to 'unify' a palette are by a)lowering contrast b)making all objects' hue's closer together c)adding shadows between objects
*shadow example: grass meets dirt, shadow on the edges, rock meets 'whatever': crevices added as an excuse to add a shadow tone
-one other way to blend tones is to blurr the hell out of the surrounding area
*blurr example: crevices near rock, tiny indistinguishable details prevent your eye on focusing on that area creating a 'blurr/blend' effect
*blur example #2: grass meets dirt: adding small patches of grass scattered in the dirt helps guide the eye into the brown easier
*because the dirt is generally lighter, you slowly 'loose' the darkest green tone, this causes the new grass to be less contrasted and thus blend easier

+green: needs to be darker still (too overpowering, green tones take awhile to unify, one of the trickier colors)
*needs to be shifted closer to yellow as well
+brown: needs to be more 'yellowish' to blend better with the grass
*since yellow is a lighter color, more contrast is needed as well
*since the contrast was added, I had to replace several tones I laid down near the rocks, since they were too dark (I used the next lightest tone)
+grey: minor contrast tweaks, saturated the darker tones to blend better with the ground
*remember when choosing tones, their colors are BASED on whatever tones are nearby (if you plan on having a decent looking image at least)

[STEP 5]

-this last step is full of subtle tweaks, and finishing off the image
-added two more dirt tones, one highlight and another detail tone
-{more to come}

[ALTERNATE EDIT]
{FINISHED!}

-wilder colors make an image more vibrant and dynamic
-contrast makes an image look more real...and dynamic
-shadows make an image look even more real and more dyamic
-you cannot just shift the hues around, darker hues (purple/red etc) need less contrast than brighter colors (yellow green etc) to look good, too much contrast will fudge up the viewer hehe
-initially you do just play with the hues...but then you must unify again (modify contrast, saturation etc)
-note that you can actually shade by using hues, darker tones use darker hues (usually) this requires less of a saturation/luminance change between colors
*removed one tone from the bottom layer to make the viewer focus on the 'cliff'
*add a light-fall-out, the further away from the viewer, the darker the scene gets (purples near the top of the scene)

I hope what I put so far is of some help, if you have any specfic questions on anything I wrote, just ask.

[Helm]

This is good, so far, although i haven't run the numbers at all. I'll post with how I work later on, some of the rules I use are different from how dhaos' rules, but the application can be close.

[Faktablad]

Thank you so much, Dhaos.  This will certainly be extremely valuable to me: I learned a lot, and I'll certainly work on it.  Thanks for going to so much trouble, I appreciate it. 

[Dhaos]

Helm: it should be really interesting to see how you setup one of your images, color-wise at least

Faktablad: I added the alt-edit (after step 5) using some more wacky coloring and shading, this concludes my tutorial-ish-thing. Enjoy.

[Krizmo]

Wow, that was a really helpful tutorial Dhaos, I sure learned a lot more from it. The other thing I would like to know about is line arting, and what you '1337' pixellers do to make it come out the way you want, I can never get that. But that probably isn't for this topic..... anyway, can't wait to hear what Helm has to say on this.

[Silver]

Helm: it should be really interesting to see how you setup one of your images, color-wise at least

I agree, if you have a free time you can take your take ..

[Conzeit]

dammit dhaos. that seriously kicks ass. the result is awesome, and I love the selfimportant-LESS tone you gave it.

With such good ground base, this might become a tut topic...and I just thought I'd voice out my opinion on tuts....take it however you like :p dismiss it if it sounds like rubbish.

I think tuts should be read and made more like marks on the path, not something to be worshiped and followed as a complete truth or fact. Tuts are about the way one person does it.
That's why I dont really enjoy the preaching tone in most....and it somewhat relates to that Alucard topic I made a while ago, I think tuts should ideally be more open like that...or like dhaos'es =)

[Helm]

Okay here goes. I won't be drawing anything as I explain as I have no time. I go into colour theory in both my TRVE MASTERS OF NEBULAR FROST thread and in my Yellow Sign self-portrait one, because I was asked for explanations on how I did my colours there.

http://www.cs.brown.edu/courses/cs092/VA10/HTML/GoethesTriangle.html

this thing above is real fun, I suggest you use it.

Now, it's still guesswork for me, but there are basic rules. I lack deep technical knowledge, but generally I avoid too much saturation in natural surfaces, ultra saturation is for plastic colour and human made reflectives, of which I don't draw much anyway. So as long as you keep saturation low, you pick a hue, any hue well whatever hue you think will make a good base. Let's say we're making a human face. A desaturated dry flesh tone will do. From there, we consider the lightsource tint. Lightsources have tints. daylight is bright yellow and reflective blues and what have you. Therefore we start making the shades go towards the yellow highlight we need. The darker shades should complement the yellow with the opposite colour in the colourwheel, in this case, purple. So darker shades mean less saturation, and towards purple. Lighter shades, more saturation (naturally, but don't overdo it). That's pretty basic, and I've done the purple-flesh-yellow skin ramp for spritework for the last 3 years or something. However, here I have 16 shades, which is way more than I'd need for modest purple-flesh-yellows. so what do I do with intermediate shades? I decide to represent other parts of the colourwheel, do near neighbour tints. Now, this is a bit guesswork and a bit 'lol let's see how much we can cram into this thing lol!' but not as much as you'd think. For example, I selected mid-range pinks/oranges because I knew I'd need them for where the flesh is pink with sanguine humour so to speak, and greens because I wanted that sort of zombie thing going on. So when you have a few primary tints, and a few secondary tints, you try to bridge from one shade to the next, the fine art of minmaxing lightness/constrast/hue to have a servicable ramp. for example, when I entered the first orange on the cheek in step 3 after pep suggested oranges, I just selected the colour below (a blue shade) and shifted the hue to orange to see how it looks. However, this is not a finalized shade. For this shade to be used for maximal effect, it has to NOT be the same lightness and saturation as the shade below it, it needs to be lighter, or darker, even if it's by a little bit ( you can get shades to work with eachother even if they're 10 or 20 lightness apart if you tweak the saturation. It's really hit and miss here ) so it becomes part of the RAMP, not just an alternate tint to be used only here and there. My focus is on using all the colours everywhere, at this time in my pixel art path, therefore unifying the palette is no1 priority on any piece for me. This creates, however, baroque-ish and monochromatic pieces. everything just melts into itself, a wonderful flurry of colours. This is the effect I want, but others prefer segmented colours and more game-arty things. Look at the three blick coloured pieces by myself, tomi and pep. Look how Tomi made it into a game sprite, with 3 ramps ( blue, red, green ) pep applied his special stylistics with the bright but unified palette, and mine is sorta monochromatic, every colour everywhere...

hope the rambling helps. If you need any clarifications, ask.

the other posts

I don't think there's any tinting way early in the process here, is there? Post which history step you mean if you want something specific. When dealing with demoscene artwork, I make my colour ramps estimating what tints I'll be using beforehand and then adjust each shade until it meshes considerably in my mind. Way early this is all the one flesh ramp until I'm satisfied with the volumetrics. The corpsepaint face of course called for a pure grayscale ramp, and the hair and leather for blues and reds. So 64 colours. When I had my ramps, I somewhat intuitively started testing tints on places ( like the collarbones which I REALLY love as they came out ) and some stuff just makes sense ( like the tint on the shoulder from the resonant blue behind the hair ). Other stuff are just there to be there. I like mixing everything with everything. Pure ramps look boring to me.

Craig Mullins is talking about things I probably wouldn't understand for the life of me I guess, but from my limited ability and experience I can too tell that as long as you properly minmax the saturation and lightness of a shade, you can do pretty freaking big jumps in hue from shade to shade and it will look very well. Look at a recent edit I did for faceless' avatar for hue jumps in a single ramp. There's theory behind the hue shifts most of the time, but since pixel art allows for such minute control at any level in creating the piece, one sometimes just goes crazy and experiments with the HSL sliders on a shade, 1 bit at a time

Generally though, the theory is that you tint towards the colour of the lightsource ( blueish yellow in sunlight I guess ) and towards complementary shades in darkness. Skin strangely adds saturation in darkness in places where the skin is thin and light subsurface scatters through, but it goes towards more muted unsaturated purple darks where it is thick and oily. In pixel art, and this is more of the way I learnt to pixel from studying amiga-era artists, I usually forget proper theory when I tint art. I like to unify my palette for the piece, everything everywhere, and I like to keep the main colour of the piece as 'ambient'. It's there, but it's not THERE and THERE but not THERE. It's the general colour. Otherwise I like to use otherworldly lightsources, tints and smearings that while not realistic, suit the pixel art aesthetic I've developed over the years.

The computer is not a tool for simple reproduction of representational art like actual painting can be (amongst other things). It has it's own aesthetic. There is an aesthetic in the method. You can accentuate the aesthetic by making the method more visible in the art. Computer art, video-game art, pixel art. These things have special charges that need to be understood and to an extent exploredfor me to respect the produced art as within a new medium in itself. In that respect then, I pixel in colour schemes and stylistics I've picked up from amiga art, because clear 'reality'-based colour theory doesn't accentuate the pixel art method for me. Also, I sometimes texture using squares, or my structures ( not as in this piece ) accentuate the vertical and horisontal grids, generally giving the feel of squares, doublewide pixels, etc etc. Pixel art isn't just drawing in another way.

about green tint in flesh: I think I was trying to unify that end of the ramp with the pure gray ramp I use to tint that under the nipple for example for later when I would collapse the two ramps so as to save shades at the optimization stage. But the sickly green suits his oily flesh, no? What else would probably work would be as sick burned pink, but it would need a lot of juggling to balance it now if I changed it.

 I'll basically reiterate a lot of the advice I gave there, here, but this time non-specific to particular pieces. I have to overstate this point however: I DO art much more than I THINK about art. In the immesurable number of minute choices one's mechanism makes when doing a piece of art, I am 'aware' and rationalize only a very small part of them. The rest happen on their own, and I do stuff I think looks good again and again until it's subconscious too. So for every 'rule' I have for myself as I'll explain it below, there's a million exceptions or stuff that might even be contradictory in practise. Every art piece calls for it's own set of definitions, and it's an exercise in frustration to try to create a realistic working model of 'how colours are chosen' that applies to EVERY piece.

This being said, colour choice for me is about 3 things in pixel art: colour conservation, palette unification and tinting. These things have a lot to do with each other, but below I will adress them pretty much seperately. Everybody can make their own connections about how one helps the other and the like. As I said, this is highly artificial and non-relevant to how art is actually made so I won't try to make it appear more natural.

COLOUR CONSERVATION:

Few colours per piece is important to me. The Computer Aesthetic ( CA for short thereafter, I will refer to this a lot) as I understand it and appreciate it is very much tied with artificial limitations. The building blocks are square (pixels) the colours are few and sometimes garish ( EGA, C64 palettes), you have to resort to tiled repetition, dithering to fake shades or texture etc. These things I believe, make pixel art it's own sort of art, and those that use the medium without appreciating and accentuating the CA are basically doing painting, only using pixels. I believe this is a disservice to the medium, and personally do not do this. If I want to paint something, I'll paint something. If I want to do pixel art, I'll do pixel art, approaching the CA deliberately and with respect.

So colour conservation makes no sense outside of CA, there's no other 'why?' to answer. We just have grown to using few colours then because of machine limitations, now because it accentuates the CA. I believe the less colours you use without loss of detail and information, the closer your are to representation of CA and therefore, the more the art you make appeals to me (personally, not talking about anyone else). I like it when my art appeals to me, therefore this is what I do. Makes sense? About detail and information. detail is pixel placements. If you can use less colours and still maintain the pixel placements that you've done, then this is good. This is relatively easy. If you take a 256 colour pic and you turn it into 128 shades of gray, it is very possible even automatically, a machine will induce no loss of detail.  Information isn't pixel placement, though, it is WHAT the pixels are. What colour, what they signify as texture and symbol. If you can use less colours and still retain the informative signifier-aspect of pixels, then you've really become a master at colour conservation. This is more difficult because information is implied in pixel art through minute shifts in hue, saturation and lightness. 3 bright blue pixels on the tip of a character's nose mean something different than 3 yellow pixels you might arrive at when you coalesce your colour ramps to keep the palette down. Therefore, it's a lost art that I personally consider very important, to juggle information, detail, against the goal of small, concise palettes. I am losing this battle in most pieces I do, because I end up with odd numbers of colours like 18 or 9 and I have this urge to JUST TAKE A FEW MORE COLOURS OUT so I can reach the canon of 16, 8. I end up with quite monochromatic pieces, which a lot of people may like, but I myself know I could have let breathe with more shades here and there.

So, I start work with grayscale 16 colours usually, taking out the first 2 and the last 3 (most pixel art doesn't need close-to-but-not-really white/black slots) so I have PURE BLACK - darkest gray, darker gray, dark gray, darkish gray, GRAY, lightish gray, light gray,  lighter gray, lightest gray - PURE WHITE, right? I fill the background with an extra shade, usually close to GRAY, or lightish gray, but tinted to a pleasing earthly tone, very little saturation, something green or brown or blue, depending what I think the sprite will be put against most of the time.

I block out the image volumetrics with GRAY, and when I'm happy with the outline, I use a darker shade of gray, maybe two steps down, to do outlines of the various parts of the sprite/image. When I am happy with this, it is time to think about colour.

Now, every piece for me has one or two predominant shades, and a couple of shades of lesser importance. In most reasonably lit pieces, GRAY, and the two steps up and two steps down from it will make most of the wide surfaces, so when I'm happy with my design, I add some saturation to those shades and play around until one major colour theme is represented. A lot of this proccess can't be quantified exactly, let's just say that I am left with a pretty pleasant not-to-saturated monochromatic sprite that washes to gray in both brighter and darker shades.

TINTING:

Now, I start to add different tints to different levels of the ramp, trying to think ahead what colour I want where. This doesn't have to be too exact because I'll be editing these a lot as I work. This goes hand in hand with me adding detail to the sprite, though not a lot of information. Still mostly doing shapes, light and dark against the main colour scheme. The theory behind tinting is still a little shaky, but I've learnt to guess what works and go with that, but always try to be a little adventurous as well. I consider the light of the lightsource, which usually is blue with yellow for daylight/sunlight, as well as secondary lightsources I might want to use (usually saturated blue or purple) and I try to move TOWARDS the tint of the lightsource as I go towards bright, and AGAINST the tint as i go towards darkness. What I mean by AGAINST, is that I go towards the opposite end of the colour wheel as I go to darkest shades.  Purple is the opposite of yellow, and so forth. This is an expressionist technique that I am not really able to back up with numbers or theory of light or anything, it just looks good for me and leaves me space to do tints I am interested in and also helps in unification later so there you go. Now, both my BRIGHTEST but not white and DARKEST but not black shades I do special things with. Usually both of them are full saturation for me, which might not make much sense in realistic terms, but I like it because a) when you're really close to white, all colours seem to the eye to be the same, the less saturation, the more this happens. So if you do a subtle tint, it'll just show as pure white. So I don't do subtlety there, I go full out and do magenta or pure yellow brightest shades for effect. These USUALLY are in the colour of the main lightsource, but some deviation occurs depending to what the rest of the shades need and are doing. This is a holistic effect, so you can't adress it in vacuum. My darkest shade should be close to gray though, realistically. Not much saturation exists in darkness. But because this is pixel art and most usually sprite art, the darkest shade is a cool place to give a secondary tint to the whole piece. If I have a brown piece, I might do FULL SATURATION 30 lightness blue for a tint or something else. Study my pieces for many examples of this. Now then from main shade, we go towards lightness tint in bright, and opposite end in darkness, correct? How do we do this? By doing steps along the way, clockwise or counterclockwise on the colour wheel. A red jacket might then benefit from a step to orange, and then a BRIGHTEST shade of loud green before settling on pure white, whereas a pair of blue pants might go the opposite way and step on cyan, then a desaturated green/brown before settling on a YELLOW highlight before pure white. This is basically all my tinting theory on this level.

When I'm done with most of the detailing of the sprite, I'm left with a haphazardly tinted piece, that needs to be unified.

UNIFICATION:

This is a pretty straightforward procedure. You go through all the shades, doing small HSL alterations to tie it all together. What this means is this: if I have two shades that look as if they're pretty much the same level but of different tint, I differentiate them a bit, making one lighter and one darker, so I can also use them as buffer shades, or to AA places. This is one of the easiest pieces of advice to follow in this text: if two shades are almost the same in lightness, make one darker, one lighter. This has advanced usages, though. Sometimes you do this differentiation but then you need the yellow you made darker to also look brighter in some other part of the piece. This is all about context. You pump more saturation into the yellow, and you settle it against less-saturated shades that might as well be brighter than the yellow, and the yellow will look brighter anyway. Such tricks you pick up as you practise pixel art.

When all the shades are ONE ramp, one lighter than the other, you might have to go through everything and pull the hue value of disparate shades a bit closer to the main hue of the piece so it's a bit cohesive. Even wildly coloured pieces like my avatar in this forum, still have main shade that everything else gravitates towards, namely blue. Squint a little: that main tone your eyes infer from the piece is the one every shade should make a small step towards.

When all this is done, I look at my ramp. Do I have any near-duplicates? What can I take out without loss of information or detail? The LESS I have to do on this stage the more it means I'm closer to controlling my palette properly. Sadly, I usually take out one or even two slots of colour at the end, and minmax the rest to make for it's loss, so this is as I said not an exact science. I usually try to use most of my colours in most of my pic, very rarely leaving detail in without graduation of information about it as well.


So there you have it, more or less. I'm not sure how much of this is helpful in any meaningful way, but if it was interesting to read at least, cool.



This is a quick edit of dhaos' excellent example to show what I'd do to the palette. 16 colours now. (note very quick and dirty bit where I use stone colours on dirt base and plant colours on dirt base and vv. Every colour everywhere, unification etc)

[Silver]

Thank you helm !