About

INTRODUCTION // This project started out with the creation of a “font machine” that generates typography based on textual input. Early prototypes of this system were mainly based on randomness (a number of yes/no questions that roughly determined the shape of a letter). Since then, the system has evolved to a rule-based algorithm that creates compositions and typography based on the content of textual input. This blog serves as a place to report my progress. The following text describes the evolution of the process itself from the start to about 1.5 year ago. More up-to-date info comming soon, but too many to-do’s for the moment.

COINCIDENCE // The evolution my font and compositions has been a course of experimentation, dismissal and recyclation. I first set out to make letters composed of elements I distilled out of photographs. But after reading about the I-Tjing I rapidly changed course towards creating a dynamic font that was totally subjected to coincidence. In other words a font in which the shape of a letter depends on chance. And I know what you’re thinking: “I-Tjing?!”. But I-Tjing was but a starting inspiration in the creation of my font and with regard to cognizability I’ll give you a short brief on I-Tjing. This way you can clearly see the origin of my starting point.

I-TJING // The I-Tjing is an ancient Chinese book that is used to predict the future and provides the user with vague advice to guide you with life questions. To consult the I-Tjing you have to reflect on the question you want to ask, while repeating it over and over in your mind. After thinking about the question for a while, you have to take 3 coins and toss them six times. Both sides of the coin each have a value attached to them, one side 2 and the other 3. After each coin toss you have to add up those three values and take note of the result: an even result corresponds to a full line and an uneven result to a broken line. The line that corresponds to each result is chronologically noted below the previous line-result. After six tosses you get a specific hexagram that correlates to a certain sign. Then you have to use a special table to find out which sign this hexagram represents in the I-Tjing. With the lines of the first three tosses you refer to the horizontal index, with the last three lines you refer to the vertical index. On the crossroads of both you’ll find a sign that represents a page in the I-Tjing and on this page you will find the answer to your question.

WHAT I USED // The thing I found interesting and inspiring about this process is that a number of random yes-no questions determine the form of a sign. I just replaced the word “sign” with “letter” to get to my initial concept. In other words: “ a number of yes-no questions determine the appearance or form of a letter”. So I used this as a starting point to begin my search for a system to realize this concept, using nodebox as the medium.

SEMI-AUTOMATIC TYPE // So how does it work? For the line grid to function – or any grid for that matter -I needed general constants, so I adopted three lengths: “long, medium and short”. With these three lengths I could then set up a general grid for letters through coding in nodebox. For example a “D” is made up out of one long line, three medium lines and two small lines. I also started to take pictures of all sorts of windows, ranging from basic houses to cars, from churches to modern architecture. Out of these pictures I copied certain windows and used those to make different lines in Photoshop. For each length I created a database consisting of 64 variations of lines. In nodebox I could then build a letter out of placeholders, in other words if there is a long line present in a letter, that long line could then be switched with any one of those 64 in the long line database. The result is that the letter changes form, because different lines are selected each time you run the process. You can compare the entire process with making letters with matchsticks. For example when you position matchsticks on a table to form an “E”, you can replace any of those matchsticks with another one, it will always resemble an “E” if you respect the position of the matchsticks. And that is what nodebox does; I can determine that position and make nodebox remember how the “matchsticks” should be positioned to resemble a letter.

1. Picture of apartment
2. Copy element out of picture through drawing
3. Copied element
4. Composed line with copied ornament
5. Database of long, medium and short lines
6. General letter structure in Nodebox
7. Filling of general grid
8. Result

The step to compositions was actually based on an early experiment with the square grid that I recycled in my line grid. The idea was more an experimental approach to type and I abandoned it before because I was focusing on creating a conventional font in which readability was still an important factor. I started from the thought that a grid letter is built up out of horizontal and vertical lines of certain well-determined lengths. With those two constants I could then break the conventional structure of a letter and let the parts roam free in the grid. This way a letter would become more of a composition than a medium for reading. I basically transferred this concept to the line grid. The lines out of which the letter is built, could position themselves freely – only respecting their orientation and length. This way letters, words and sentences would become compositions. From this point on my fascination for the word compositions grew and my focus shifted from conventional typography to these experimental language representations. The fact that sentences would have a counterpart composition, the dark industrial and abstract feel and later on the notion of a controlled chaos kept me fascinated.

CATEGORY // From very early on I had been playing with the idea of working with themes, something I also recycled into my new process, combined with the idea of connecting the content of the text with the form of the compositions and type. I wanted to create a system where the type and compositions would relate to the content of the text. I wanted to be able to describe a certain place and get type and compositions that relate to that description. Of course not in the literal sense that typing the word “elephant” would result in a font built out of elephants. But rather a font built out of lines that symbolically represent the nature of the subject. The symbolical shape of these lines would be subjected to what I thought these themed lines ought to look like. To consequently apply this gigantic content related principle, I divided the world into 4 categories: motion, architecture, nature and objects. The shape of the lines would vary depending on the content. Each theme or category would have its own characteristic database that would represent the theme in question. All lines would also be subjected to a specific theme-related coordinate structure. This coordinate structure would be controlled by the motion category. With the 4 categories or themes I could then set up a theme oriented dictionary in nodebox. This way the process could recognize any word that corresponds to one of the themes and would then subject it to the relevant database and coordinate structure.

MOTION // The motion category would consist out of 3 variations of dynamic coordinate structures, an invisible category that directs the three other categories and determines the position of every line. It represents the invisible structures that dominate observation such as time, point of view and motion.  In the process, motion is a coordinate structure that becomes more complex every time a word of a certain theme is used. Each of the other three categories (i.e. nature, architecture and objects) has a different coordinate structure. For architecture this coordinate structure is built up out of parallel vertical lines that start at the bottom of the page. Each time an architectural word is used a new line appears. The nature category coordinates are built out of a series of ovals that grow in size with each nature word. In the object category the coordinates are based on a dynamic path built out of connecting horizontal and vertical lines. With each “object word” a new horizontal or vertical line appears at the ending of the previous line.

BASIC // All of these coordinate structures are off course still very basic and should only be considered as a starting point. In the near future I will also be implementing a new category that represents abstract content. This category will most likely formally represent exact geometrical shapes.

LINE SHAPE // Concerning the form of the category lines I first wanted to repeat the process used in the previous type and compositions. In other words taking photographs and copying elements out of them through drawing. But I quickly realized that the three categories needed contrast to clearly reflect content, otherwise the compositions would become one mass of similar objects. So I used specific keywords as a starting point and formal guideline.

THEME-CONTENT RELATION // The architectural category applies to all words that have something to do with buildings, houses, windows, or in other words, obviously, everything that has something to do with architecture. The formal keywords for this category were perspective and firm. The lines themselves would have a strictly horizontal and vertical orientation. The nature category represented all auna and flora. The keywords were free and amorphous; the line orientation would have no restrictions. The object category represented all sorts of objects like cars, garbage, clutter, … The keywords of the object category were documenting and objective. For orientation the lines would be restricted to 0°, 45°, 90°, 135°, 180°, 225°, 270°, 315° and 360°.

There have been much more developments though. This text describes the startingpoint and basic core, describing SATCIC from 2009 to 2010. So it’s really in need for an update, but I’m pressed for time at the moment (coming soon eh).