Fledgling designers and typography geeks can now get a helping hand with picking the perfect font thanks to a site that uses machine learning.
FontJoy will select pairings from a database of more than 1,800 fonts when the user hits the ‘generate button’.
Finding a selection of contrasting yet complementary typefaces can be a challenge, but this clever site makes it as easy as possible.
The site includes options for headline, introduction and body text fonts and the simple interface enables the user to toggle between different combinations.
They can ’lock’ one or more while the site flips through suggestions for the remaining ones, or leave all three unlocked for a slightly more flexible approach.
Alternatively, users can choose each font manually.
The site is the work of Jack Qiao who, according to his website, hesitates to call himself either a designer or an engineer, but has experience in both fields.
To create the site, Qiao extracted vectors from images of fonts - these are used to encode the visual information of each letter.
This enabled him to use vector arithmetic to create a neural net that compares different typefaces and find those that pair well together.
"Through vector arithmetic we can isolate the features that represent visual concepts like serifs, obliqueness, and weight - or even more abstract concepts like legibility, kerning and colour", explains Qiao on github.
"One of the more interesting things we can do with font vectors is font pairing, or the problem of selecting fonts that work together in a design".
While such methods are unlikely to replace experienced human designers anytime soon, the site is a useful tool for newbies or just a bit of fun for typography nerds.
This article was originally published by WIRED UK
