Pixelation Knowledge Repository

By PDLCPayday Loan

by Helm of Pixelation forums

Introduction - Back to the top

Pixel art is like a Go game. Every pixel placed contributes towards a struggle between intentions that will eventually end in an state of variable equilibrium between the opposing forces. Anyone that has played Go knows of the feeling of immense mental strain when they place one piece down on the board and they ponder on the eventual ramifications of that single move towards all the clusters of pieces across the board. A piece feeds on the vital space around it, you see. Not all pixel clusters can live, some will have to die for the purposes of other, more important - towards the whole of the game - clusters. Go is a holistic game, a game about sacrifice and a game about grace. Likewise, drawing with pixels results to similar concerns of balance between clusters of pixels and their optimal positions are again about sacrifice and grace.

Let's consider what a cluster of pixel is as opposed to a single pixel.

This is a single pixel: .
This is a cluster of pixels: _ or that: ,

The cluster of pixels is made from single pixels. However, a single pixel is most of the time near-useless and meaningless if not touching pixels of the same color.  
The pixel artist is concerned with the shapes that occur when pixels of similar color touch each other and convey an opaque, flat, shape.
Most of the defeats and possible triumphs of pixel art occur in that exact moment where the artist makes a cluster of pixels.

What are the defining characteristics of a cluster of pixels on the morphological level? Besides those obvious and common with other types of art (like the information inherent to the color in terms of value/brightness, chroma/saturation or hue/tint that the cluster possesses) we are interested in the particular characteristics of that body of color as pixels. The characteristics of the shape are defined by its outline. It could be made out of straight lines, 'perfect' angled lines (will return to perfect lines later), implied curves or irregular (or jaggy) lines. A cluster has often many of these attributes around its outline. The prime job of the pixel artist is to find the ideal shape for every cluster while considering how they all come together to represent the item they are rendering. It is my belief that there are almost ideal shapes for clusters of pixels and they are those that achieve a twofold, yet holistic goal: how to optimize the resolution of the image. "What's this?" I hear you say "but isn't the resolution set anyway?" The real resolution of the image certainly is. But the fake, that is, the perceived resolution of an image is in the hands of the capable pixel artist, higher than the real one.

There it becomes important to realize what the available fineness of resolution exists for the piece of art the artist is trying to render. The less colors the artist has to convey his image, the more the available resolution tends towards 1 pixel = 1 pixel. The more colors the artist has, the more they can approximate, fake essentially, higher resolutions by proper buffering. Look at this image:


The 256 color gradient makes it impossible to even notice a pixel. The effective resolution here is considerably larger than if the artist had 4 indexes to convey the range between black and white.

In pixel art we do not deal with 256 color gradients, however and therefore the effect isn't anything bigger than perhaps 1 real pixel = 0.75 fake pixels, but it is still a very important thing to consider and makes or breaks great pieces. When looking at a piece of pixel art, the artist should be able to evaluate how many colors can be used to blend clusters of pixels better so as to optimize towards a finer resolution. This is not a point of stylistics, this is what pixels long to be, that is their ideal form. Any style can benefit from this process.

A beginning pixel artist should always start with very constrained palettes, where hue and saturation do not matter, just value. Gameboy 4 colors is excellent. Small sizes, small palette. 1 pixel = 1 pixel there and they can worry more about dealing with how clusters of pixels long for their perfect shapes together rather than care about anti-alias or dithering and other advanced resolution-upping techniques. If you can't render your item with 4 colors in a gameboy screen, you will not be able to do any better at 800x600 with 256 colors or more.

What is the perfect shape of a cluster? It has to do with its outline. The juggling act here is to think of what you're trying to represent with your pixels and then try to retain its essence while at the same time making the clusters you're using to draw it become as close as they can to 'perfect' lines. Perfect lines are the 90 and 45 degree ones. Curves can be assembled from smaller segments of perfect angles also. Avoid single pixel noise. Using perfect lines, before the artist even starts to anti-alias manually, the contours of his clusters should be as close as they can be to 1 pixel = 1 pixel resolution. 


A is a freehand doodle line of pixels. This is an implied cluster too, even if it's a line. B is the same line, cleaned manually until it's made of the safest couplings of pixels possible while still retaining the intended curvature. 

On the detail of A we see that the jaggies hurt the resolution of the image by conveying larger pixels than our computer monitors are capable of displaying. This is effectively, the bane of pixel art. Banding does this. Bad AA does this. Bad dithering does this. Pillow shading does this.

While perfect vertical and horisontal lines are an amazing help to the pixel artist that is trying to refine their resolution as much as possible in some ways, they do the exact opposite than what they intend to in others


Here in A we have Ideal clusters of black and yellow. The resolution is practically infinite. Then on B we have two lines of buffer between the two colors. The artist might have thought they were making their image softer and therefore the resolution finer (a correct principle generally) but what they've done instead is completely give away the actual dimension of their smallest-possible pixel line. Pixel drawing is about the art of obfuscation of the single pixel. The more you can hide the actual dimension of the prime building block of the image by presenting perfect clusters of pixels in harmony, the finer the resolution of the image becomes. Of course it's impossible to completely avoid the side-products of buffering, but this principle should be kept in mind regardless.

Banding is horrible partly because of the above explained principle and also due to the 'breaks' in the bands that are ever-so amplified the more the artist piles on bands, effectively lowering the resolution of his image

As mentioned before, the goal of the pixel craftsman is twofold. On one hand he tries to make the apparent resolution finer. On the other he struggles to represent what it is he's drawing. The two goals are always in friction. Let's look at these two pieces of high art I just devised:


And this:


I posit that both images represent the same idea. A man shitting in a sine arch.

The first image has a high apparent resolution because the lines are perfect and also - more importantly - because as the viewer looks at this they cannot discern a pixel grid, they cannot see the single pixel almost at all.

In the second image the single pixel is very apparent. If we zoom in a bit more in fact,


That there is a very confined space in which to represent the human figure in its volumes and shapes means that the more colors and clusters we employ, the more the underlying grid of the image has to show.

The particular ambitions of the pixel artist, at this level, are paramount. If one wishes to convey a fully rendered object as realistically as possible, it cannot be helped that the apparent resolution will be lowered. As clusters of pixels come to interact, places where the pixels 'line up' and betray their resolution are inevitable. The trade-off is that the rendered object appears more realistic, with whatever benefits the artist might assign to that. Let's look at a schema:


Here we have a simplified model of aesthetic motivation for the pixel artist. It isn't very different from the motivations of artists in other fields, but there are some additional considerations to inspect that are very pixel-art relevant.  On this point I'd like to say that I am not using the above terms as they're usually meant in the history of art. I am appropriating the terms slightly so the layman can follow along. 

With abstraction I mean that the realized object of the piece of art does not clearly refer to something in the natural world.
With realism I mean that the artist is attempting to render his object lushly enough that the viewer will interpret it relatively literally.
With symbolism I mean that the artist is attempting to convey objects with clear higher functions without allowing for literal interpretations.

Try to think of your favourite pixel artists, and place them in a space within the triangle. Try to assign specific works by them in the triangle and then make specific observations about how each artist is prioritizing their two goals: hiding the grid/increasing apparent resolution and conveying the volume, light and surface of their intended object literally. You will find that artists near the top of the triangle will have very high resolutions and very simple/naive objects, whereas artists near the left end of the triangle will have resolutions of moderate fineness while their objects will be meticulously shaded. Art towards the right edge of the triangle will both have very low resolutions and very simple shapes!

The realist pixel artist will often make large areas where the resolution is practically infinite (like the shoulder of the girl in the above Lazur bit) and then place single-pixel, low-resolution level detail on various specific pieces to rejoice in the pixel-ness of his work just a little (the highlights on the hair here for example). The ambition of the artist leans heavily towards removing the grid, but doesn't frown away from going 'hey, here's my pixel, do you love it? I love it!' once in a while.

The complete abstractionist has effectively destroyed the pixel in his work, it is in the place of Ideal Space. It could be vector art or anything else that isn't shackled to the limitations of a monitor really. We do not have examples of such pixel artists really because as you might imagine, that goal would be very self-defeating. However there are a few artists whose work is very very close to vector smooth, like Panda or Ilkke sometimes, but you can tell they're pixel artists at heart because they can't contain themselves from putting in pixel-level details in a few places after all.

The symbolist pixel artist creates art that is very informed about its being made of pixels and wants the viewer to know it also. All of the modern 'retro art' fits in that edge of the triangle for me, with the artificially low resolutions and the flat and fat pixel character designs. These retro artists are not interested in pixel art technique to make the resolution higher, they are interested in invoking nostalgia on the older viewer or to inform the younger viewer of the semiotic particulars of older video game art. 

I do not judge any of these motivations. It is however my belief that regardless of which way the artist might feel drawn towards, for their art to maximize its capacity as pixel art, they should reconcile their different aspirations so as to retain a place within the relative center of the triangle (the grey circle area). The realist artist should not attempt to completely abolish the pixel-level detail and end up with a blurry mess of a piece with 250 colors in it. The symbolist artist should not completely forego the attempt to make their pixel clusters achieve their ideal state. The abstractionist should not make their resolution that fine so that in the single pixel no longer feels like it belongs.

Banding Primer - Back to the top

If you use pixels to draw something, it is inevitable that some banding will occur. Banding is when pixel clusters 'hug' each other to create visible resolution lowering. Since the pixel artist will be putting clusters of pixels close to each other in order to communicate certain shapes that are one way and not any other, it is inevitable that they will run against situations where they must create minor banding to convey succesfully what they're drawing. This is a risk that the artist should be willing to take. However there occurs a lot of banding by the inexperienced artist (or the experienced artist in a hurry) that has no artistic merit. Let's look at art by an experienced pixel artist that seemed to be a bit in a hurry that is to his benefit to go over and eradicate. This from the amazing videogame Spelunky which I've been playing a lot as of late and as such I've had the time to look at its art a lot and I've spotted the bandings. The artist is Derek Yu

(note I do not use my own art to spot banding because I can no longer create STRONG banding by mistake, it sticks out like a sore thumb to me and if I draw it up on purpose it seems disingenuous. Also note that the following examination is in the interest of learning and not meant at all as a disrespect for the artist.)


Here is a play-screen of Spelunky. The art is pretty beautiful in my opinion, but it shows signs of 'working fast'. This makes sense as Spelunky is still in beta form, not finalized and Derek did what any smart artist would do, he created assets to put in the game and left time for fixing the little things at the end of the project. It is a very useful circumstance for us however because we can go in and look at it in it's intermmediate shape and examine the methodology that creates common errors such as banding. Let's zoom in and check the main sprite and the basic rock tile next to him.


In red I have blocked the places where (I submit) unintentional banding occurs. In purple I have outlined further cases of banding that are not fatal (because they are artistic compromises made by the artist in order for him to convey the shapes he intends to) but are best addressed also. As you can see the unifying characteristic of banding is that it conveys a lower resolution than intended. The viewer spots fat pixels, or rows of fat pixels. 

In the attempt to move around pixels in such a tight little place, until the banding is resolved, the design points of the character move around a lot, whereas the rocks next to him are very easily rid of the banding without many changes. Why does this occur? It occurs because the character has a higher semiotic value for the viewer, because he has a face, and arms, and a torso, and all these things must be communicated clearly, whereas a bunch of rocks are just a bunch of rocks and there's not much problem just moving around a few pixels here and there. But the whole demeanor of the character sprite can change by a single pixel's worth of alteration. I posit this argument: 

1. the more semiotically charged what the artist is trying to rid of banding is, the more costly in terms of characterization single pixel changes are.
2. the smaller (in pixels) the size of the item the artist is trying to rid of banding is, the more costly in terms of characterization the single pixel changes are


3. The smallest the character sprite is in resolution, the more difficult it is to remove banding while retaining it's intended characteristics.

Proof of concept: Go change a single pixel on the pacman open mouth frame. Or a single pixel from a NES final fantasy battle sprite. Just a single one. See how much the idea of the character changes. But change a single pixel in a 64x64 sprite. Not much difference. Furthermore, change a single pixel in an 64x64 ground tile. Almost no emotional impact to the change. 

Consider here the informative qualities of a single pixel in the different contexts. If a pixel is supposed to inform the viewer of a character's eye, it is extremely important. If a single pixel is supposed to convey that this is a bit of shading on a rock, not so important. This ties with the Information versus Detail qualities of pixels, something to write about more later on. 

Anyway, here I deal with the banding on that sprite:


I urge the viewer to open the animation in their animation program and inspect every change closely. Smaller changes around the banding that seemingly make no sense will start to once the viewer starts to see how every change impacts the areas around it holistically. How removal of banding also triggers the impulse to alter and fix manual anti-alias. This process, is what refining ones pixel art entails. This, in my opinion is, what it means to take something composed of pixels, and make it pixel art. This is the zen stage of pixel art, where antialias, dithering, banding-removal and other minute handlings of pixels bestows on the piece the blessings and limitations of pixel art. Not everything made of pixels is pixel art. Not even all art made of pixels is pixel art. The full immersion in that state of definition and refinement outlined briefly above usually and optimally, creates results that are very much great pixel art. It's not a binary switch "PIXEL ART / NOT PIXEL ART", it's a gradual slope. The more time the artist spends in that zen state of minutely switching around placements of pixel clusters, the more his end result will be graced by the benefits of great pixel art.

Let's also look at a bigger piece, again from Spelunky. This is part of the intro splash screen:


Nice and big so banding can be spotted. Here's a challenge. Go over it in your pixel art program and spot the banding just on the Indy character in the middle. Then check below for my take.


Here's most, but not all, important banding errors in this part of the image (extra challenge! spot the additional minor banding errors! I can see at least 3 more). They show a few different types of banding. Let's go over them:

A, H :  Just basic 'Fat Pixel' banding. An illusionary effect of where 4 real pixels convey a shape that seems like a double-res pixel. It is very important to note that the Fat Pixel, when used intentionally and intuitively creates specific effects that are very controlled and worthwhile. It does remain a lowering of the resolution but sometimes that's exactly what the pixel artist needs.

B: Row of Fat Pixels. This is the worst, especially when there's a row followed by a row just a pixel displaced lower or higher. I mean this:


Or even worse, stuff like this:


This is BANDING THE ULTIMATE (or more handily I dub it 'staircase banding') and it makes my heart bleed. Obviously these are not spotted in the Spelunky art because wherea Derek Yu might or might not be very theoretically aware of the banding effect as discussed in this text, like any good pixel artist, he certailny must find gross errors like the above aesthetically insulting even on an artistic-instinctual level.

C: Skipping-One Banding. When an additional cluster is banding with a section that isn't directly touching it. This isn't a big problem, but once the eye is trained to spot these alignments, they cannot be unseen.

D, E: This is an amazing bit of banding. Check how It's two Fat Pixels, Skipping-One row to create a Row of Fat Pixels. I show this to explain to the reader that they should train their eyes to not only notice banding on immediate pixel rows, but even 'higher level' banding. Essentially, the viewer must learn to project horisontal, vertical, and 45 degree lines from every piece of banding cluster in their art and see how it effects the pixels in these directions all the way to the edges of their piece. This sort of training makes the artist holistically aware of how their piece is structured also, and this is a great artistic skill to have regardless of pixels. (Should I write a Banding 202 article here for this or is it beyond the scope of this tutorial?)

F: This is the most common type of banding. It's when a full outline of a shape is hugged by close shading. There's much worse examples than this in most amateur pixel art (like the awful staircases I showed above) but Derek is not an amateur by any chance so it's not a very pronounced example. Check below for how when fixing it this takes a lot of extensive reworking and aa of the shape.

G: This is banding as well. 45 degree lines band pretty horribly in fact.

Here's some suggestions for fixes:


Again, I suggest the reader takes this in their animation program and checks the choices made to remove banding.


I stress that some degree of banding is inescapable in pixel art. Don't obsess over erasing every little bit of banding all the time, but keep in mind that the most glaring errors effectively destroy the illusion of increased perspective on your art and most importantly (since a lot of symbolist pixel artists simply don't care about hiding the actual resolution/ increasing the effective resolution) banding betrays sloppiness. Why? Because when someone draws in a non-digital medium, where there is no grid,they can go from general shape to specific detail without creating visual effects such as banding. But on the computer screen, when the artist places a general, rough shape, and then later on puts another cluster of pixels somewhere closeby, it is very probable that exactly because of the small resolution, some edges of these clusters will have lined up, creating banding. When this ends up in finalized pixel art, it tells the discerning viewer that the artist didn't spend the time to optimally place every pixel cluster in his image. And that is a big part of what makes pixel art a potent artistic medium.

Pixel Clusters - Back to the top

Let's look at this sprite from a NES game from 1992, Little Samson:


In the loop I show the clusters that make up his face, headband and hair.

A cluster of pixels of the same color here, exactly because the NES could allow only 3 colors (plus one for transparency) for the sprite, becomes extremely important. The headband is just one, the hair is just one, the shape of the face is one. The eye is one, and the other eye is another. This is a very pure example. 

If we change a single pixel of these important clusters, what they signify changes, along with their relationship with the clusters around them. 
Here's what happens when we move a single pixel around on each frame in various places. Note how the balances and relationships between elements change


NES art is a very useful step in learning to do pixel art because it forces the beginner pixel artist to realize the power of pixel clusters and finding ideal shapes for them to work in unison to convey the intended characterization, without getting bogged down in dozens of colors and a huge resolution.

Let's look at something a bit more complicated:


This is a piece by the lovely Kenneth Fejer. You can see his work here.

Let's look at a detail of the piece up close:


The cluster of pixels outlined in red is where we aim our attention. Around it there is another color tone that is halfway between the bright green and the dark green. Single pixels smooth out the pixel cluster we're looking at. That is called buffering or manual anti-aliasing. I will not go into detail as to what antialiasing is here, more capable artists/german-aa-machines are working on texts to fulfill this purpose. Let's for now assume you are well aware of how anti-aliasing works. The reason I am showing you this piece in particular is because if you look at the buffer shade there between the two main colors, you'd think 'well... these pixels aren't a pixel cluster, are they'. They do not touch, they are - mostly - single pixels, so what are they exactly? 

They are part of the meta-cluster that they are buffering towards, or from. For reasons of simplicity, it is best to think of them as always belonging to the smaller cluster that is touching the bigger cluster, but this doesn't really matter in applicative terms. 

Once the artist realizes that the anti-aliasing around a cluster is nothing more than part of that cluster and its main purpose is to define that shape better, they stand to expel a lot of the anxiety and confusion that usually occurs to the beginner when they're faced with the near-infinite options of pixel placement that are available when they start a new piece. The artist can stop concerning themselves with just placing single pixels willy-nilly and replacing them and rearranging them by trial and error "until something looks right" and can instead apply a functional, progressive way in which to build their art. Pixel clusters are the tool that the pixel artist uses to convey 99% of the render of their object,not pixels themselves. The beginning artist should start a piece with a single pixel cluster shaped as the silluette of what they're trying to convey. Let's say, for example that I want to draw a face.


This is what I start with. Then the silhouette should be segmented to temporary planes. The experienced artist, once comfortable with visualizing his model might not have to do this stage and go directly to the next one, but for the purposes of this tutorial I'll go ahead and do it:


This is more or less a simplified 3d wireframe, with the planes of the face. This isn't lit yet, it just shows me a selection of shapes from which on the very next step I shall pick to make my pixel clusters. Naturally the smaller the piece the more difficult to actually pixel this stage, but it's not difficult to imagine it, and the artist should always imagine the factual planes of what they're attempting to render.


This is really the most important level in establishing ones pixel clusters. I have chosen a light-source and lit some of the planes. They have created pixel clusters of different colors that are competing in the small space for definition and information. Please note that I have not anti-aliased consciously, nor have I removed much of the apparent banding yet. If your pixel art doesn't look solid in this stage, no amount of pixel polish and tricks later on will save it. This is where your traditional skills come into play. Note also that the palette I'm using isn't very contrasted yet, this doesn't matter. If you light the planes correctly, you can then adjust the contrast as much as you want the the piece will still stand. 

Here for example is a drastic contrast adjustment via fast remapping:

"The horror... The horror."

Look at the planes in this stage before we move on. Isn't it easy to tell where buffering should be applied? It's a matter of common sense. The sharper the edge, the less anti-alias needed. The smoother the transition, the more anti-aliasing will be needed. The buffer pixels should never overpower the cluster they belong to. 


Here I have refined and antialiased the clusters into metaclusters. Keep in mind that Anti-aliasing can create banding! Look below:


If the buffering pixels line up with an edge below they will band. This is an extremely common error with anti-alias-happy artists and they can spot that something looks 'off' but can't put their finger on what it is exactly. Well, now you know. This is where a pixel artist expert in anti-aliasing shows their true skill. This is the thick of the fray, as it were. Adjusting single buffer pixels until they don't band, but yet represent the intended shapes. 


Here I do more and introduce dithering. Not a lot of it is needed in most pieces of pixel art. Dithering isn't a mystery of any sort, think of it as the tapered, fading edge of a brush stroke... the dither belongs to a parent cluster. The places where one needs to dither following this methodology are self-evident, it's where I didn't have enough colors to make a transition smoother. Again this creates banding (dither-aa-clash) and it renegotiates some unclaimed space from the flat pixel clusters. Solving these problems elegantly is what pixel artistry is about. 


Though color is beyond the scope of this tutorial, here's the final piece with a palette. It's very easy to colorize a solid grayscale construction. Then bits of banding around the outline were removed and a final refinement and it's done.

Dithering - Back to the top
Basic Dithering

On the bare level (which is the one the artist should approach this effect most of the time) dithering is like buffering. It belongs to a pixel cluster and serves the end of smoothing parts of it which touch other pixel clusters. Let us consider this schema:


The top form is an 1bit pixel cluster. Think of this as an oily brush stroke. If the artist just has the two colors to work with yet they wanted the shape to taper off or be softer in places they could employ dithering very selectively like in the bottom form. Look how the dithering emphasizes the flow of the pixel cluster and most importantly look how the dithering doesn't dominate the cluster. If there is more dithering belonging to a cluster than there is solid, opaque color, then the dithering becomes a cluster in itself. This is a necessary function for dithering some times, especially in environments where the artist may use only very few colors to convey objects that have a lot of levels. But for the purposes of a basic survey of the effect, the artist is encouraged to keep the dithering as part of the cluster it belongs to and to use it to emphasize the flow of the form.

The middle form is a slightly different proposition. Here the artist has more shades. He has used them to break up the brush stroke cluster into smaller clusters of pixels. Note how if you squint, the middle form and the form below it are very similar. Dithering = buffering. However, in segmenting the cluster into smaller clusters then the artist has the opportunity to smooth things out further within the meta-cluster by dithering between the available colors, as seen in the right, and final, form. Again the artist should keep in mind the flow of the meta-cluster and now additionally the flow of the smaller clusters that it comprises of. In most cases the pixel artist will not be working with such large clusters. Instead they may find themselves wrestling with 3-4 pixels at a time trying to find where they belong. This is the essence of pixel art, finding where each pixel belongs in relation to the clusters that are vying for its inclusion. In small spaces, the power of the single pixel, as we've discussed, becomes stronger. A single pixel can make or break a cluster and therefore dithering serves to confuse the forms more than it smooths them. Let's then consider this argument:

1. The smaller the pixel space the artist has to work with, the less dithering they should employ for purposes of smoothing between clusters.
2. The more colors the artist has to work with, the less dithering they need to employ for the same purpose.
      3. In most pieces of small, unrestricted (in terms of color count) pixel art don't be surprised if you never have to dither for purposes of smoothing. 
(there might be a call for dithering for textural purposes, to be explained later)

{Furthermore this would explain why dithering is almost never employed in the rendering of video game sprites, where as we've mentioned before, besides usually being small, also show single pixels as being very powerful in conveying specific information.}

If the viewer zooms in on the example they will note that for this I used the 50% checkerboard dither, and then the 25% version where every second line (vertical or horizontal, it's the same) is omitted. The reason for this is very simple: When I desire to do basic dithering, I start with 50% dither and establish a good cluster shape that augments the flow I am looking for. Then I ease the edges of this 50% pattern more into the clusters they are bridging towards with the 25% dither, and then finally some spare pixels around that. This ensure that the priority of shapes, as they take place in the image is always Solid Color > 50% dither > 25% dither. Again, in most basic applications of dithering, the artist will not be called to use 25% dithering (or other, less busy patterns) in big surfaces and as clusters in themselves. This fundamental exercise comes in handy much more than it restricts the beginning artist.

Let's look at what not to do:


Here we have a hierarchy of errors. The initial cluster is ill-defined. This occurs a lot with beginning pixel artists that start by tracing some reference (like a photo of their girlfriend) at a very high resolution and then they place clusters without any flow or optimization. Then they try to, effectivel, blur between the bad clusters, creating amorphous shapes without priority. When you have a bad metacluster, dithering will only serve to make it worse. Especially if the dithering is overdone and it dominates the initial shapes. What beginners then often end up with is not Pixel Art, but a piece of art made of pixels.