Computational models
for expressive
dimensional typography
by
Peter Sungil Cho

Thesis proposal
for the degree of Master of Science
at the Massachusetts Institute of Technology
October 1998

Thesis Advisor
John Maeda
Sony Career Development Assistant Professor
of Media Arts and Sciences
Assistant Professor of Design and Computation
MIT Media Laboratory

Thesis Reader
Daniel Boyarski
Professor and Director of Graduate Studies
School of Design
Carnegie Mellon University

Thesis Reader
Matthew Carter
Principal
Carter & Cone Inc.


Abstract

This thesis proposal explores the prospect of typographic forms, based on new computational models, which can be faithfully realized only in a three-dimensional, interactive environment. These new models allow for manipulation of letterform attributes including representation, scale, two-dimensional structure and three-dimensional sculptural form.

In this proposed research, each computational model must accomodate the variation in letter shapes, while trying to balance flexibility in manipulation and functionality with the beauty and legibility of fine typography.


Table of Contents

Motivation 4
Background 5
Experiments 8
Proposed Research 10
Proposed Accomplishments 12
Time Frame and Working Procedure 12
Resources 13
References 14
Readers 15

But, taking the shape of A to be that which the judgement of the mind lays down, we have to conform it to the nature of the machine and not to attempt to impose upon mechanical production either those ornamental exuberances which are natural and proper enough to human beings working with their hands, or those peculiarities of detail which are proper to the pen, the chisel, and the graver.
Eric Gill, An Essay on Typography, 1931


Motivation

We are at a point in time when the fields of art and design are being transformed by technology. In many cases, technology influences the artistic domains through the introduction of new tools which speed the established work process. In other cases, technology allows the artist or designer to create something which would not be possible with conventional methods. Some of the current work along these lines, such as colorized electron microscope photographs of scientific components (Frankel) and photograph mosaics which from a distance reveal a larger photographic image (Silvers), represents a new bent or high-tech method (one might say, gimmick) applied to an established field in the visual arts. Other work, however, represents an informed attempt to use technology from the ground up to create an unexpected, innovative, and intelligent visual experience.

This thesis research attempts to expand the field of typography through an informed use of computation. Typography, or the arrangement of text elements, is an art, practiced and explored through book, publication, and poster design throughout the past few hundred years, and in film title design in the past few decades. Type design, too, is an art with a tradition which extends to the beginning of written communication. With the advent of the web and the rise of the information age, we have seen an explosion in "new" typography: degenerate typefaces, letters which float, fly, scream, melt, decay, or otherwise animate as they try to communicate their message. Now is the time to evaluate the computational models which are used to represent the shapes we read as letters on the computer display and to look at new expressive ways these models can be manipulated.

By building up new computational representations for typography, we can examine new ways letterforms can exist in a virtual three-dimensional environment. We can explore new ways that typeforms can be displayed and new behaviors they can take on. We can also see how these custom representations affect the way a message can be conveyed and even experiment with new kinds of messages made possible or more accessible through a new representation.

The intention is that the lessons learned through this research will become more relevant as the computational medium matures. The concepts explored here can be applied to problems in information and interaction design as they become increasingly complex.

This proposal discusses the prospect of typography based on new underlying computational models which allow for the manipulation of three-dimensional and two-dimensional typographic forms, beginning at the level of the single character and moving up to the level of a word, a phrase, and beyond.


Background

Type has evolved throughout history, at times carved into clay, chiseled into stone, and cast into metal. Today, our predominant use of type consists of going into the font menu on the computer desktop to choose a Helvetica Bold or Times New Roman. We use type every day without thinking about its underlying representation. On the computer, type is represented as a list of two-dimensional data points connected by straight and curved segments which form the outline of a shape we read as a letter (or any other character).

Several different computational models of typography have been introduced in the past few decades, including the two basic schemes we use today: TrueType and Postscript. In both of these font languages, each character is defined precisely as a list of points. A type designer using TrueType or Postscript painstakingly positions the points which comprise a set of contours which are filled, creating a positive form. In general, these data points can be accessed by users of the typeface only through vector-based drawing software tools which allow the type to be converted to modifiable paths.

From 1979 to 1986, programming guru Donald Knuth developed Metafont, a computational type design system based on parameters (Knuth, 1986). With Metafont, a user can define a complete font of characters by specifying sixty-two parameters such as x-height, width, stroke weight, and serif qualities. This system represents a departure from the labor-intensive, rigid definition of type used by TrueType or Postscript. In a sense this is a more "intelligent" approach to defining a computational model of typography because Metafont has pre-coded information about what a generic alphabet should look like then uses the parameters to create a specific set of characters. While extensive and quite clever, the system has some drawbacks: it does not allow for personality in individual characters, and it perhaps suffers from the fact that Knuth, as he readily admits, is not a type designer (Knuth, p.vi). In spite of the system's shortcomings, the typefaces created by Metafont are certainly adequate and are available to users of the typesetting system Tex, also developed by Knuth.

Emigre, a type house and magazine founded by Zuzana Licko and Rudy VanderLans, represents a typographic movement spurred by the introduction and proliferation of the Apple Macintosh computer in the 1980s (Licko, 1998). Since the tools at first did not allow for high-resolution typography, Licko and VanderLans embraced the highly pixelated graphic nature of early Macintosh fonts and developed their own abstracted and geometrically ornate typefaces. As the tools emerged, type on the computer could "do more" and the interest in type as high-tech image led to a movement to use technology to obliterate type beyond recognition. The work of David Carson in RayGun and other trend-setting publications proclaimed "the end of type," with its use of typography to subvert the message.

Along the lines of Emigre's experimental letterforms, some attempts to encode computational variation into typefaces have been made within the scope of the Postscript standard. Just van Rossum and Erik van Blokland's font Beowolf, for example, uses computational randomness to create slightly different instances of characters every time they are printed. Van Blokland's font Nimida, similarly, randomly degenerates its letterforms.

Also computational in nature, but much more evocative of a designer's intention, John Maeda's work represents an enormous inspiration for this thesis research. His work, including the Reactive Books series, brings the issue of "computational expression" to the forefront. In several pieces, namely Flying Letters, 12 o'clocks, and Tap, Type, Write, he challenges our perception of how the computer can react to and perform with the user and how typography can be an elemental part of that interaction.

Prof. Maeda offered a MIT Media Laboratory course in Digital Typography in the fall of 1997. This course explored many issues about type on the computer including, though not limited to, different computational representations for type design. Tom White, a fellow student in the Aesthetics and Computation Group, developed an alphabet in which letterforms were drawn by the path of a bouncing ball in a physical system. He designed the letters by assigning a dozen parameters, including starting position and velocity of the ball, the strength of gravity and the drag of the system, then set the ball loose. This is an example of a playful and truly unique computational model for type.

Much of the formal research into expressive temporal typography also comes from the MIT Media Laboratory. The work of David Small (1987) and Yin Yin Wong (1995) in particular has contributed to the concept of typography living in a three-dimensional space. Small applied simulated Newtonian physics to typography, treating type on the screen as if it were part of the physical world. Wong developed a framework for thinking about and designing temporal typography, along with some compelling examples. Also the work of Suguru Ishizaki (1996), and his comparison of typographic elements to dancers in a choreographed piece, has been informative to this thesis research.

Finally, this discussion of typography would not be complete without a mention of pre-computer examples of expressive typography which are an inspiration for this thesis research. Examples of dynamic and expressive typography can be found in Bauhaus works. Lazlo Moholy-Nagy's poster for Pneumatik tires, for instance, gives letterforms a sense of motion by distorting the type and placing it on a plane other than the plane of the page. One of the many examples of temporal typography from motion picture titles comes from Saul Bass. In his opening titles for Psycho, for instance, horizontal lines move across the picture and reveal broken letterforms, conveying uneasiness and alluding to the jarring violence of the shower scene. This thesis would not be possible without the revolution in the use of typography in design over the past century.


Experiments

The majority of typographic experiments I have conducted thus far fall into three categories: intra-letter shape manipulations; inter-letter transitions; and three-dimensional spatial explorations. This section describes these experiments and suggests some new directions of research.

Intra-letter shape manipulations. Several of the early experiments treat letters as malleable shapes that can move in fluid ways. In the first piece, a single letterform, the A, was developed as a playful interactive element. As the user moves the mouse, the letter moves and changes shape, appearing to dance and smile. In other early experiments, entire words were treated as malleable shapes which deform in expressive ways. In two pieces, Sleepy and Aim, the words are defined by outline points which are manipulated on the screen by a script or the mouse. In both of these experiments, the points move toward or away from a single point in a manner which gives a fluid quality to the emergent letterform shapes. In another piece, Oh, the letterforms are defined by a list of points that make up the skeleton of the letter shape, along with a designated stroke width at each point. In a sense, the letters in this piece are drawn by a variable width pen. Here, the points are again gently manipulated by a script or the mouse to give an evocative quality to the letters.

Inter-letter transitions. Another set of typographic experiments are devoted to creating transitions between different letters. In each of these exercises, type is represented by a computational model which allows for an algorithmic transformation from one letter shape to another. This model can vary greatly in its visual style, from abstract pie-piece letters to three-dimensional block letters to outline sans serif letters. In each case, the user inputs different letters through the keyboard or other device and the onscreen form transitions smoothly from letter to letter.

Three-dimensional spatial explorations. The third set of experiments are loosely based around the idea of how letterforms can be used in an exploration of three-dimensional space. In each of these typographic compositions, a text was positioned in space into a sculptural form. The goals were two-fold: first, to explore the implied three-dimensional space, treating it in the same way a graphic designer usually approaches the two-dimensional space of the page; and second, to use the third dimension to reflect the text content in some way. A poem by Samuel Keyser entitled "The Crooked God," for example, is displayed as a rotating sign post of straight and bent text lines. Amy Bloom's short story, "Love is Not a Pie," is displayed in space to illustrate which parts of the story are dialogue and which are monologue. A page from the Media Lab web site becomes an abstract spiraling structure, anchored by a monolithic block composed of the names of Media Lab members. These experiments explore the spatial aspects of type in a virtual three-dimensional environment and prepare the way for new ways type can be realized in implied space.

A different exploration of type in space, Forefont is a system for modelling letters based on the idea of metaballs, or iso-surfaces created by point sources. This rudimentary experiment demonstrates a novel way of representing and visualizing letter forms.

An important part of the research conducted thus far involves developing applications which showcase the type model's functionality in a compelling manner. In Typerphonic, for instance, each letter appears on the screen through a transition which is mapped to a phoneme sound triggered by a key press. The transition (the M, for example, begins as a light, narrow letter, then transitions to a fat, wide letter) is made possible through the underlying parametric representation of type. Similarly, the interaction in Letterspace, gestures in space being mapped to different letters, demonstrates a coupling between the fluid visual letter transitions and the smooth physical motions. The aim of this research is to develop both the underlying algorithms for representing type and the applications, either expressive or functional or some combination of the two, that utilize the capabilities of the model.


Proposed Research

I plan to continue research on computational models of typography and software applications which make use of these models.

The research into typographic models involves two main aspects. The first is to develop more typefaces based on different concepts. Some of these include a parametric model, a combined outline and skeletal representation, and a model based on three-dimensional flow fields. I would like to explore these ideas both in a programming environment and in a commercial modelling environment such as Houdini.

The second aspect involves developing the basic text building block object in a functioning platform, namely acWorlds, an environment based on OpenGL currently being developed for use in the MIT Media Lab Aesthetics and Computation Group. This text object will incorporate the lessons I have learned from using both "real" and custom type in three-dimensional space. The text object should be scaleable to allow for use at a wide range of levels, from a single character to a large body of text. The text object needs to allow for basic editing of typographic attributes such as justification, tracking, and leading. The object also needs to take into account the computational requirements of having many texture-mapped characters on the screen. I also plan to provide a framework for developing and using custom typefaces which may be represented in a way different from the customary anti-aliased character texture map. Above all, this text object needs to be more "self-aware," encoded with information about itself, in a way which is thus far unclear.

Along with developing the underlying computational models for type, I am developing the applications which demonstrate the use of the type models. I propose two applications, one at a small scale of a few dozen words, and one at a larger scale of a few thousand characters.

At the smaller scale is the expressive poem. Referring to modernist typography, which flourished in the 1910s and 20s, El Lissitzky asserted that "the new book demands new writers" (Spencer, p.9). One could argue that, similarly, the new computational medium demands new writers and new messages. I propose to write and develop a poem which relies on computation for its message to be communicated. This may involve developing a model for the type which allows for words to be embedded within other words or for type to "spawn" other type. This functionality could allow for a different kind of message, storytelling, or wayfinding.

On the larger scale of a few thousand characters is a rudimentary word processor. I would like to develop a prototype for an expressive text editing environment based on a new typographic model, realized in a three-dimensional graphics space. This system should allow for conventional input of text along with different ways of visualizing and manipulating the type. One area of interest may be the ability to see the user's history of text input and editing. The experiments I have done with spatial visualization of type should be informative for this system.


Proposed Accomplishments

The deliverables for this thesis research include: a short history of type in motion, an extensible text object within the acWorlds environment, approximately eight different type models, and two applications using one or more of these models, namely, an expressive, interactive poem and a text editing environment prototype.

Since this is a thesis conducted within a technical environment, the MIT Media Lab, the issue comes up: how will I evaluate the success of this work? I propose that this research be evaluated empirically through informal critique sessions with my thesis readers, collegues at the Media Lab, and visitors. The work can be evaluated in the context of two criteria: how well the work demonstrates a computational model for type which allows for an expressive use not possible with conventional type, and how effective the piece is overall.


Time Frame and Working Procedure

November: Develop computational models for typefaces which allow for expressive manipulation of letterforms. Research the history of typography in motion, particularly in film. Begin planning text object within the acWorlds environment.

December: Plan, implement and revise text object in acWorlds. Create expressive poem application using custom type. Present poster session at Media Lab.

Jan+February: Experiment with developing sculptural letterforms using Houdini or other three-dimensional modelling software system. Continue to develop more type models. Work on text editor or other visualization of dynamic medium- to large-sized data set using custom and traditional typography.

March: Bring projects to a close. Compile a simple demo station of projects to date. Document and test applications through peer review. Begin writing thesis document.

April: Write and revise thesis document.

May 8, 1999: Final thesis document due.


Resources

I will be working primarily in the Performer and Open GL environments on the SGI Octane and Onyx machines within our research group. I will also use Houdini, a software package available on the SGIs. Most of the interaction I am interested in can be achieved through the mouse and keyboard, though I would like to try other devices such as the Flock of Birds or Polhemus magnetic field positional sensor. I have access to both of these devices at the Media Laboratory.

References

Cho, Peter. "Pliant Type: Development and Temporal Manipulation of Expressive, Malleable Typography." Bachelors thesis, Massachusetts Institute of Technology, 1997.

Frankel, Felice and Whitesides, George. On the Surface of Things. Chronicle Books, San Francisco. 1997.

Ionesco, Eugene., Massin, and Cohen. The Bald Soprano. Grove Press. 1956.

Ishizaki, Suguru. "Typographic Performance: Continuous Design Solutions as Emergent Behaviors of Active Agents." Doctoral thesis, Massachusetts Institute of Technology, 1996.

Knuth, Donald. Computer Modern Typefaces. Addison Wesley, Reading, Massachusetts. 1986.

Lewis, Jason. "Dynamic Poetry: Introductory Remarks to a Digital Medium." Masters thesis, Royal College of Art, 1996.

Licko, Zuzana., Poyner, VanderLans, and Wild. Emigre. Rosbeek, The Netherlands. 1998.

Maeda, John. Flying Letters. Digitalogue, Tokyo. 1996.

Miller, J. Abbott. Dimensional Typography. Kiosk. 1996.

Shamir, Ariel and Rappoport, Ari. "Feature-Based Design of Fonts Using Constraints." International Conference on Raster Imaging and Digital Typography. 1998.

Silvers, Rob. Photomosaics. Henry Holt and Company. 1997.

Small, David. "Expressive Typography: High Quality Dynamic and Responsive Typography in the Electronic Environment." Masters thesis, Massachusetts Institute of Technology, 1987.

Spencer, Herbert. Pioneers of Modern Typography. Lund Humphries, London. 1969.

Strassmann, Steve. "Hairy Brushes." ACM 20, no. 4 (August 1986): pp 225-232.

Wong, Yin Yin. "Temporal Typography: Characterization of Time-Varying Typographic Forms." Masters thesis, Massachusetts Institute of Technology, 1995.


Readers

Daniel Boyarski is a Professor and Director of Graduate Studies in the School of Design at Carnegie Mellon University. He teaches courses in typography, information design, and human-computer interaction design and is interested in how words, images, motion, and sound work together to produce effective communication pieces. Externally, he lectures at conferences and consults with companies on interaction design issues. B.A., St. John's University; M.F.A., Indiana University; post-graduate work at Allgemeine Gewerbeschule Basel, Switzerland.

Matthew Carter was trained in the traditional crafts of typefounding; in the forty years since then he has designed and made type in metal, film and electrons. Before starting Carter & Cone Type with Cherie Cone in 1991, he had worked for Linotype and for Bitstream. He is a Royal Designer for Industry and a recipient of the Chrysler Award for Innovation in Design, the Type Directors Club Medal and the AIGA Medal.