Days in the Typographic Labs
Three days with 280 of the frontmost people in the business of type design and technology
TYPO Labs 2017 in Berlin had 28 sessions and workshops focused around new font technology, typography experiments, “not-yet-possible” ideas and suggestions to solutions of problems of various language scripts. Ever since the announcement of the OpenType 1.8 specification (now 1.8.1) in September 2016 most online discussions and experiments have focused around the new “Font Variations”.
Font Variations
When installing a typeface, you typically install several font files. One for Regular, one for Regular Italic, one for Bold and one for Bold Italic. Font Variations eliminates the need for multiple files, as you only store one font that can be interpolated to different weights and styles. It may include “instances” that suggest what Bold and Regular is, but you can choose to interpolate it only half the way and create your own SemiBold – right in the app where you are using it (InDesign, Word …).
The obvious uses for Font Variations are weight and width: Interpolating between Light and Bold, Condensed and Extended. But the possibilities are endless. Weight and width are only two axes, and the format supports up to 64.000 of them, and the font designer can specify what should be interpolatable, and how the axis should be named. A commonly used example of the creative possibilities of variable fonts is the Decovar font by David Berlow, that introduces no less than 15 axes to manipulate serifs, stems, skeleton and much more. An entire design system that the user can tinker with.
Decovar is a modular parametric display font with a variety of skeletons and terminals designed by David Berlow for Google. It is free and open source and can be accessed from fontbureau.typenetwork.com. Illustration: Font Bureau
What is it good for?
The excitement over the new possibilities with font variations or variable fonts (the terminology seems as unstandardised as the specifications themselves) was also evident at TYPO Labs 2017 where 70% (best guess) of the speaking time covered this particular subject, from all possible angles: How to convert legacy fonts? How to harmonise outlines? How can it be used? How should it be used? What does the business model look like? When will we be able to use it in our desktop publishing apps?
Despite the broad coverage of the topic, answers to many of these questions remain unclear. Tools necessary for developing variations are already available, but the specification may still change. Great user interfaces for dealing with the variations are yet to be demonstrated, as we still only have web interfaces and demo apps simply providing sliders for each axis. But imagine that the end-user has to navigate 15 sliders, all affecting each other, to manipulate or “design” a font variation perfect for the project. A serious challenge lies ahead for user experience designers.
Artificial Neural Networks
Another interesting trend in typography, which also surfaced in many sessions at TYPO Labs 2017 was artificial neural networks. How can we train machines to lessen the manual labor in font production?
Artificial neural networks are quite easy to explain, but extremely difficult to understand. Learning from biology how neurons in our brain communicate and make decisions, this is replicated in computers. By connecting artificial neurons in huge networks, supplying them with training data, it can learn and apply its newfound “knowledge” to new data. After showing the network 10.000 photos of a horse, and telling the network that what it can see is a horse, it will no matter what new photo you show it, be able to tell if there is a horse on the photo.
Neural networks are often used for image recognition and machine translation, but can it be applied to typography as well? Some of the big type foundries are already using it, to automatically categorise huge libraries of typefaces. This can be very useful for designers looking for the perfect typeface for a project (e.g. a modern slab serif face that express 60% authority and 30% gentleness and 10% gracefulness).
But other experiments have also been made. For example to automatically generate a full alphabet from only designing 4 letters, or generating completely new typefaces from nothing. Decreasing the workload of spacing (kerning) character pairs is a likely future use case for neural networks that will save hundreds of hours in the production workflow of new typefaces.
Utilising neural networks can be very useful for designers looking for the perfect typeface for a project, e.g. a modern slab serif face that express 60% authority and 30% gentleness and 10% gracefulness.
Takeaways
Even though text has been around “forever”, font technology is obviously not a static thing. We are excited about the not-so-distant future developments, and look forward to follow and contribute in what we see as attainable goals for the font technology industry in the coming years:
Decrease font file size (for web)
Ease workflow, e.g. by better automatic spacing and hinting (for type designers)
Enhance support for non-latin scripts (for the masses)
Develop user interfaces for accessing font variation axes – that aids the designer creating more beautiful typography
Create tools that aid designers choosing typefaces
The 2nd TYPO Labs Conference was held in Berlin, April 6–8, 2017. TYPO Labs an annual summit on font technology and engineering.
The conference brings together leading engineers and practitioners from the font industry, OS developers and academics. Its purpose is to advance the state of the art in type development, facilitate its adoption in the industry, and encourage the integration of new type developments into future digital communication.
Since more and more software and devices rely on written communication, the need to synchronise technical efforts is now more urgent than ever.
Video recordings of the all the sessions can be found at here.