How Fontjoy Generates Font Pairings

Updated

Fontjoy generates font pairings by running font combinations through a neural network trained to recognize pairings that share an overarching theme while keeping a pleasing contrast. You click Generate to get a new pairing, Lock to keep fonts you like, and Edit to choose a font manually. The site describes the result as "font pairing made simple" and points to its GitHub page for the technical details behind the model.

What the tool actually does

Fontjoy is a font-pairing generator. Its stated goal is to help designers pick combinations of fonts that work together. The page frames pairing as a design judgment — which fonts work together is "largely a matter of intuition" — and then applies a neural net to that problem instead of leaving it entirely to trial and error.

The pairing principle it follows is explicit: choose fonts that share an overarching theme yet have a pleasing contrast. That is the criterion the model is built around, so a generated pair is meant to be related enough to feel coherent and different enough to create visual interest.

How a session works

The page gives three controls, each with a clear effect:

Control Action What you get
Generate Click to create a new font pairing A fresh pair of fonts
Lock Lock fonts you want to keep The locked font stays fixed while the other changes
Edit Choose a font manually You set a font yourself instead of accepting the generated one

The preview text is editable. The page suggests replacing it with your company name or other copy, which lets you judge the pairing on realistic content rather than placeholder text.

A practical loop follows from these controls: generate until one font looks right, lock it, then keep generating to find a partner for it. If the model's suggestions miss, Edit lets you override a slot directly.

Why a neural net, and where to look closer

Fontjoy's own explanation is short: the problem is approached with a neural net, and more technical detail lives on its GitHub page. The page also links to a font visualizer and a logo maker as related tools, and offers an occasional newsletter about Fontjoy and deep learning.

For anyone evaluating the method rather than just using it, the GitHub link is the place to check — the site itself does not publish the model architecture, training data, or evaluation criteria on the main page.

What to keep in mind

  • The output is a suggestion shaped by the theme-plus-contrast rule, not a guarantee of good typography in your specific layout.
  • Locking is the main way to steer results toward a font you have already committed to.
  • Testing with your own copy is part of the intended workflow, since the preview text is editable.
  • The page does not state pricing or account requirements; check the site directly before relying on it for ongoing work.