Hello,

I'm new here! I'm originally a machine learning engineer and recently got interested in type design.

I wanted to link these worlds together by building a model for filling fonts & generative fonts. My focus is to make it a tool for designers (similar to what Co-pilot Cursor is doing for programmers). I'm curious if people in this group are interested in this project - what challenges you foresee, what features you think would be useful to you.

As I see it, the first thing to consider would be how the model is trained.

Probably the last thing type design need is more opportunities for font piracy and licence circumvention…

Zeit: 8.6.2025 um 15:49, Nova schrieb:

My focus is to make it a tool for designers …

The designers using fonts or the designers making fonts?

15 hours ago, Ralf Herrmann said:

The designers using fonts or the designers making fonts?

The designer making fonts. My idea is it will work similar to a copilot/gpt user flow for font designers.

As mentioned in this recent article, there is a lot of repetitive work in type design, that AI could help with. One major problem I see is the interface between our design tools and the AI tools. I can easily see AI being integrated into a font editor, doing these repetitive tasks. I am struggling to see how a standalone service could work well to aid professional type designers.

Because the designs are broken down a lot. We have a light drawing of a glyph and a black drawing to interpolate between them. Accents are always components to be placed onto base glyphs. Even parts of a letter (corners, serifs) can be reusable components, so they can be edited globally.

If the AI system doesn’t integrate with all that, it wouldn’t be very interesting for professional type designers. I’m sure it can crank out hundreds of crappy free fonts based on prompts in minutes, but I wouldn’t cheer for that.

That's an interesting article, and it seems I've got similar ideas.
I totally agree that the workflow needs to be smooth, and have considered building an extension for either Glyphs or Robofont (don't think it's wise for me to build a full font designer from scratch). Componentising is definitely in scope, especially since I want to support CJK fonts too.
Thanks for your comments and do let me know if you have any more thoughts!

One particular area where AI can kick into the type design in a useful and meaningful way is the kerning. If done properly, that would build the trust of the type design community in using and following AI tools. Since it is a laborious process and currently there are not many options to automate it (not to dive deeper here into analysis of available options).

So far, we have seen amateurish attempts to replace the whole font production process that reflect poor type design knowledge of those AI enthusiasts. Those attempts are mostly oriented toward users who are not familiar with the deep understanding of the OpenType format and the level of precision involved in professional type design.

Really cool idea! A font generation tool like that could save designers tons of time when experimenting with styles. I think the big challenges will be keeping the output usable and making it simple enough so non-tech designers can jump right in. Definitely something I’d be interested in seeing develop :D