Let's consider strictly technical aspects of icon creation
1. Lines, angles
One of the vital elements of the composition is the outline of the object, i.e. solid line, which limits the object from the rest of the image. When creating normal-sized graphics, the artist rarely cares about defining objects with complementary outlines. This is unnecessary because of the scale: even non-contrasted objects do not blur into a single whole. Icon graphics is other story. Two solutions are possible: either foreground and background colors must be from different sides of the color chart, or the object must be divided from the background by visible outlines and shadows. I will discuss lines in more detail.
If we think again to the large scale images, we see that in order to define edges, we can use all (even the most complicated) angles, Bezier curves and edges. Anyway, the line will look perfect due to anti-aliasing. When decreasing the graphic dimensions to the small icon size, situation changes dramatically. When every single pixel is vitally important and can ruin the overall look of the composition, anti-aliasing is simply unreasonable. It comes that you have to consider the possible line slopes.
The angle you select for the line, specifies the step of this line. This means that each line is made of basic elements, the combination of which defines its neatness and esthetics.
For example, a 18-degree angle contains many small 3-pixel lines combined with a 1-pixel downshift:
This is how almost each line looks like. Basic element with joining and another primary element. Unfortunately, not each angle makes the line look neet and not messy.
For example, here are 25 and 20-degree angles:
It can be easily seen which line is more attractive. The 25-degree angle produses a line containing equal 2-pixel primary elements. The 20-degree slope has a basic element containing of three parts: 3-pixel, 3-pixel and 2-pixel. The angle is more complicated, therefore, when not supported by additional visual effects, it creates a feeling of messiness: the basic elements have contrast which is too obvious to hide from the human sight.
Here we can see the first rule. The most "correct" angles must make as plain primary elements as they can. Thus, the ideal angles are 0 and 90 degrees. In them, 1-pixel primary elements are joined without downshift and produce even lines. Less ideal are slopes which form primary elements of 1, 2, 3 pixel with 1-pixel downshift. Surely, the lines can not be even, but the human sight will process the image and present to the user what you meant to show to him. Also, the following angles can be considered correct (but at a stretch): angles with basic elements consisting of different lines shifted by more than 1 pixel.
It is the time to move to another issue. In the last paragraph, I purposely defined angles with basic elements of 1, 2 and 3 pixels as acceptable. Such tiny lines, especially when a great number of them is combined, appear as a single whole. But what happens if we change the primary elements to 10 or 20 pixels?
Here is an example:
It can be obviously seen that the line loses its solidness. It is not a single line anymore; it is a combination of several lines, located near each other. Hardly any artist wants its creation to look that way. So, we come to another conclusion: if you use small angles, which make long primary elements, it has to be reasoned and used with maximum caution.
And, finally, the last thing I have to tell about lines. I did write that the basic element should be as simple as possible.
For instance, a 25-degree angle can be drawn in the two (sure more of them exists) following ways:
In the first case the basic element is a 2-pixel line. In the second picture, it consists of three elements: 3-pixel, 2-pixel and 1-pixel. The same angle but the look is different. The complex primary element produced the line untidy.
Such examples can be made for about any slope, so in your projects, try to simplify as much as possible.
It can be said without an overstatement that the color is the leading element of your work; not suiting or badly matched colors can kill even the most creative idea. What can we say about the color in miniature graphics?
First, there is a certain rule: if the image looks bad when reduced to grayscale, then the colors that were used are incorrect. This rule is valid not exclusively for icons but for the whole graphic design.
Second, you will be unable to use common RGB. Sure, you may select internet colors only, but this limits you greatly. Gradients, shadows, dithering and many other effects will not be accessible to you.
Third, try to increase the portion of "flat" color. The more consistently colored areas without blurring and jumps your project has, the more neat it appears. Pictograms are too small and the over usage of special filters makes the image look messy.
Fourth, gradients. This is a great instrument which can change the appearance of any composition. Besides visuals, gradients are a perfect solution to abolish the "broken line" look (lines with too long basic elements). But be attentive when applying them since too many gradients can kill the plain color, and be unable to fit into the gif palette.
Fifth, shadows and highlights. All rules for applying shades and flares in icon design are totally identical to the rules of the general graphic design. The only thing I would like to mention: create everything by hand. Don't apply filters, make all shades and highlights in a separate layer and then edit its transparency. When using filters, you almost never know what the result will be. It is unfavorable when you don't control the making of a composition. It is awful if you can't control it if creating miniatures.
And, lastly, sixth. Nuancing is the main aspect of icon creation, which greatly results the visual look of the project. When the green human eye in the icon is sized 1 pixel, increase the color saturation. For example, instead of 0, 131, 159 make 71, 195, 242. It is indifferent that the human eye can't have such color; in the first example it will appear like a gray dot, but in the second one it will really be a green eye.
Thoroughly control the picture, do not let the optical tricks to ruin your design. It is OK if objects is not perfect technically, but the entire composition must be perfect.
As an example, I drew the packet of juice located before me. The picture has plain color, gradients with broken lines, shadows and highlights and nuances.
If I wrote about something bigger than icon creation, this chapter wouldn't fit into a whole book, but the resolution of compositions ;limits fonts too.
The resolution of symbols is the main issue as opposed to their look. Nearly in any case, unless the letters are the primary part of the composition, the font size has to be reduced to the largest extent possible.
In general, virtually any graphic company has its unique font with tiny letters. This font can be made in a few hours. You can search the web and collect your own collection of very nice fonts. First of all, I would recommend you to download the DS Pixel and Seventen 7Vedi webpage and the complete set of fonts from the Lakmus Lab website.
If you don't want a new font, little sizes of Arial and Verdana may work. As a last resort, you can create the necessary letters by hand.
There are quite a few rules there. First, letters cannot be less than 5 pixels high or less than 3 pixels wide.
Second, letters should not use anti-aliasing. It transforms the text into an incomprehensible mixture of pixels.
Though, there are no rules without exceptions and the 5-point Arial looks rather readable.