Font classification is hard. As you probably know, a font file can be almost empty, including nothing except the font name and, well, the font glyphs themselves. Getting the information needed to classify such font as sans-serif, serif or bold, can be really tricky or almost impossible. Some fonts may include this data, but these cases are rather rare and even if a font file has this information, it most likely is inconsistent.

Many of you were asking us to add the ability to search fonts by font type, so we needed to create a robust way to determine which font belongs to a certain class. We decided to tackle the problem in two ways: training an AI to classify fonts and adding crowd tags. First, let's talk about our AI (his name is Sparky), which we have trained using machine learning.

Power of the machine. To teach a machine to classify fonts, we needed a simple scheme of font classification: what classes are primary, what subclasses they have, what classes contadict each other. Not to reinvent the wheel, we researched what is already out there, and in the end came up with this:

Then, we collected lots of font images and tagged them manually and according to our classification system. This was a lot of work, but eventually it was worth it: after our algorithm analyzed the images, it became pretty decent at classifying fonts by itself. Now it just needs a bit more testing and eventually we will run it to automatically tag all the font data in our system, and that's a lot — around 1.7 million right now.

Soon you'll see tags appearing on your fonts, and gradually, as our AI trains more, all fonts will be tagged with basic tags. But this is only the first part, now let's get to the other.

Power of the crowd. Machines are not perfect. People are much better at recognizing patterns and determining visual hierarchy and seeing a bigger picture out of smaller parts. This means they can classify fonts much better! The negative side is that they can't do it as fast as machines, but by combining the efforts of many people, a lot of great things can be achieved.

This is where you come into play. With crowd tags, every tagged font counts: when the same tag is added to the same font by different people three or more times, it becomes a crowd tag. Crowd tags are then displayed and added to fonts automatically for all FontBase users. And don't worry, you personal tags are not stored anywhere, we have a list of white-listed tags, and only if one of them is added, it is stored and tracked in our database.

Adding tags is simple: just right-click a font family or a single font and select Manage tags in the menu. All tags for currently selected fonts will be displayed and you'll be able to add new or delete existing ones (you can't delete crowd tags though).

We've also added the ability to search fonts by tags. By entering for example tags: serif, bold in the search field and pressing Enter, FontBase will display all your bold serif fonts in the currently selected collection, folder or provider. Search works the same for your personal tags and crowd tags.

We really hope that with the help of the machine and with your support, we will be able to establish the biggest and the most detailed database of classified fonts ever created. Every tag counts towards the common goal, and every FontBase user can contribute to the community!