Graphic design, form- and image-synthesis, Jannis Maroscheck © 2018–21. Mail
Order Book
Shape Grammars
published July 2020 by slanted publishers

12 systems
150.000 shapes
836 pages

42 EUR

How can unique pieces be mass produced? Or: How can the computer take over and support creative work? Sol LeWitt writes in his Sentences on Conceptual Art: “The idea becomes a machine that makes the art. [...] There are many side effects that the artist cannot imagine. These may be used as ideas for new works.” A form is removed from the status of pure art as soon as it is filled with unambiguous information or applied utility. Its poetic function as art is thus weakened, its practical function as design is strengthened.
With the right system, an idea can also become a machine that produces design instead of art. This is then called generative design. However, this form of design is primarily used to display complex data sets or to fire off overwhelming visual spectacles.

Based on the work of Sol LeWitt, graphic designer Jannis Maroscheck has designed and programmed his own production systems that can draw an unlimited number of individual graphic shapes.

The result is a systematic catalog—a kind of dictionary of shapes—for browsing and exploring geometric systems, in which one can always discover something new. Shape Grammars is intended as a handbook for graphic designers for the design of fonts, logos and pictograms, which, in addition to 150,000 generated shapes, shows some potentials and limitations of generative design. At the same time, the work serves as a basis for further research on more complex systems and artificial intelligence. The computer can thus already function as a dialog partner in the creative process.
Latent Arrangements, 2020

A big dataset of posters was collected and fed into a Generative Adversarial Network – a type of neural network, that learns to produce samples, which are similar, but specifically different from the input images.
The generated compositions are like a machine’s vague idea of a poster. They don’t carry any precise meaning, because the network doesn’t look at graphic compositions as a set of elements like word, picture, texture and symbol, that we use to communicate. Instead it follows a purely pixel-based approach to creating form, which is closer to that of painting with pigment and brush.
However, the big advantage of machine learning over traditional programming methods becomes visible. It enables generative Systems to evolve in more autonomous and surprising ways. Complex sets of design rules for hierarchy, contrast and balance do not have to be understood and defined manually by a human. Instead they can be learned from a dataset.
Phantombilder, 2018

A set of portraits, that were reconstructed from memory using 3DMM Software.
Vague Copies, 2020

Classic subjects like the still life are tackled through photogrammetry.



Jannis Maroscheck is a German graphic designer living in Tokyo. He studied at the Faculty of Design Würzburg and the Tokyo University of the Arts. He designs visual identities, print, digital and spatial products. His research practice deals with automation processes in graphic design, form- and image synthesis.