To people who do not work with the medium themselves, 'pixel art' can mean many things. Interestingly, even among pixel artists, it will mean different things. However, we have a tendency to have a much more strict view of what can and cannot be considered pixel art.
Nonetheless, for the purpose of understanding different viewpoints, here's my own (which I mostly picked up from other pixel artists).
What is and isn't pixel art?
It's not just blocky-looking art. Some pixel art out there doesn't look blocky at all. It generally refers to art made with aliased tools, but most importantly, with heavy usage pencil tool or equivalent, as found in any version of MS Paint. However, there are many kinds of styles possible with just a simple tool, and not everything made with aliased tools can be considered pixel art by the very artists who work with the medium.
It can be argued that going back to the early roots of pixel art could help us better understand the medium.
Consider what I think is the most famous NES sprites out there. I think most of us know who he is. While it's not the oldest pixel sprite out there, the NES had a significant amount of limitations - three colors per sprite, and a tiny restrictive resolution which heavily influenced character design.
Look at how they conveyed information with these heavy restrictions. The hat is a blob of red. So is the overalls. The eyes are just a two-pixel column. Where pixels of similar colors touch each other, they become a cluster. In pixel art, pixel clusters are used to convey information. Single pixels could also do the same but they are often not as effective as pixel clusters.
With that in mind, here's a comparison between pixel art, and down-sampled art:
First, the original to the left is/was an icon used by =Jynzo
. It was not meant to be called 'pixel art', but nonetheless scaling it up 300% can help show us an aesthetic difference between pixel art and down-sampled art. In a sense, 'pixel art' is a bit of a misnomer. It's not art with individual pixels, it's art with pixel clusters. 'Pixel art' made by downsampling an image to the point where individual pixels are visible will not have the same aesthetic that makes pixel art really shine.
However, more discerning viewers might look at the single pixels on the neck and wonder "the heck are you doing there, not following your own advice?" Here's the gist of what those pixels are trying to do:
On the left, we have the original, actual pixels. On the right, is what those pixels are intended to convey.
The green pixels are buffer pixels. They are strategically placed so it 'feels' like half of it belongs to the dark cluster where the other half belongs to the white cluster. If those green pixels are darker, then more of that pixel feels like it belongs on the darker cluster. This trick is called antialiasing (often shortened to AA), and in pixel art it's done manually. It may seem counterintuitive ("Hey, it's PIXEL art what are you doing? It's art... with PIXELS!").
Visible pixels isn't necessarily what makes pixel art shine (neither is dithering). Some people like seeing pixels. To some, it can be seen as lackluster craftsmanship especially if not executed most skillfully. Square blocks do not exist in nature so one can see that needlessly visible pixels loses realism. Pixel art is intrinsically sharp so even *skillfully* AA'd art is sharper than automatically AA'd works. Palette control, animation potential, sharpness, and its huge potential for game graphics are some things that definitely makes this medium shine.
Antialiasing: something that is definitely a part of pixel art.
If we zoom into the forehead of the original, what sticks out to me is that there is no clear intent on each minute change of color between pixels. I see no merit in changing this random pixel a wee bit lighter than another pixel. It is much easier to color the forehead a solid color.
Pixels are also aren't great at conveying gradients at small scales. Where two colors border, a line is created (some people may handily refer to this effect of creating lines called banding). Sometimes, creating visual effects like these is not desirable. The image of the super-zoomed forehead makes a pixel grid very visible by having unecessary color changes via creating a number of lines.
Here, I used a computer algorithm (with bad artistic reasoning) to reduce the number of colors to 11. While 9 colors were used on my own pixelled transformation of this pic, it convinces me that down-sampling images still gets us results inferior to human-made art. Though I definitely changed the light source when transforming the original for artistic effect, I still made better use of fewer colors.
Even if we put aside the process of making pixel art and focused only on the end product, we will notice that we will get inferior results if we ignore the process by taking computer shortcuts.
However, color count is not the only thing that separates pixel art from non-pixel art.
Here's a controversial accident of mine:
I apparently needed an avatar at a huge size of 128x128 but I wasn't interested in a super-detailed style. I wind up creating a faux-vector. Interestingly, this has a color count of 13. However, (hoping our monitors are similar enough) people will probably look at this and say "oh, it's a vector". With that in mind, color count is not the only thing that defines pixel art. I believe resolution (and how you use it) is just as important as color count regarding pixel art. A high color count suggests 'cheating tools' and easily creates visual effects of banding we probably do not want.
The above is something I made at a relatively large resolution but also with a simple style. The pixel cluster sizes here are enormous that pixel-level polishing techniques will not make much of an impact, if it is reasonable to apply any to begin with. Pixel artists often refer to this sort of aesthetic shown above as an oekaki. This does not have the same aesthetic as say, this following pixel art of mine:
Pixel art is the medium that puts the most consideration of the very pixel grid used to display an image. The larger we work, the proportionally smaller the pixel, the more irrelevant the pixel grid, the less there is of a point of hand-pixelling the image. If we're using pixels to make oekaki-like images, it's definitely worth considering other media that requires less grunt-work.
Consider the above image. Does it look smooth to you? To a lot of people, I would imagine the answer is yes. It's easy to say that we cannot achieve a smoother circle with pixels (duh, how do you make smooth circles with square blocks?). However, once we compare that circle to a smoother circle in this pic: [here]
, some of us may be surprised that this 'smooth' circle CAN be smoother if we applied a bit of technique.
Pixels ideally need a lot of care and attention. It's very easy to create very uneven curves that are needlessly jaggy. If we're not interested in going the extra mile to put this sort of care into our pixels, we will not be able to have the most refined results.
Giving ourselves a large canvas also gives us more pixels to juggle - the above 50x50 icon already has 2,500 pixels. Both the faux-vector and the oekaki have
a total of 16,384 pixels. By working smaller, we are also saving time and making it much easier to create a super-polished pixel art aesthetic.
Making something big while still maintaining a polished pixel art aesthetic is not a trivial task.
Ideally, pixel art with a clear pixel art aesthetic should be made of pixel clusters of a size small enough that pixel-polish techniques such as smoothing out curves (with or without AA) and dithering becomes useful and is used. Too many unpolished, large clusters risks creating an oekaki aesthetic. Use of computer shortcuts like blending, soft brushes, or automatic scaling/rotating without using nearest-neighbor interpolation essentially turns our pixel-works into mixed-media art.
Of course, there is no black-and-white. There are plenty of works of many different shades of gray on our scale from definitely-pixel art, and not-pixel art. The gray areas are likely areas which people will disagree on the most, but the things that makes some pieces fall into extremes on this scale are useful to keep in consideration.
Even if some may disagree with ideals mentioned here, I see that pixel art is a fuzzy term which is often not very well understood. Hopefully, I have offered some interesting viewpoints here.