Evolutionary Typesetting: An Automatic Approach Towards the Generation of Typographic Posters from Tweets

Abstract

The recent developments on Artificial Intelligence are expanding the tools, methods, media, and production processes on Graphic Design. Poster designs are no exception. In this paper, we present a web system that generates letterpress-inspired typographic posters using, as content, tweets posted online. The proposed system employs Natural Language Understanding approaches to recognise the emotions, the sentiments, and the colours associated with the content. Also, the system employs an Evolutionary Computation approach to generate and evolve a population of poster designs. The outputs are evaluated according to their legibility, aesthetics, and semantics, throughout an automatic fitness assignment hybrid scheme that combines a hardwired fitness function part with a multi-objective optimisation approach part. We experimented with the system to perceive its behaviour and its ability to evolve posters from contents with distinct textual purposes and lengths.