What AI tools does the font community need?

Options

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

I'm currently building Fontweaver, an AI tool for type designers. It can currently generate simple fonts! I'm curious if people in this space are interested in this project and what features you think would be useful to you. I'm eyeing:
- Kerning support
- CJK support
- Font filling (font generation based on drawn glyphs)

Tagged:

Comments

  • PabloImpallari
    PabloImpallari Posts: 821
    The current most requested feature is language extension.
  • Thomas Phinney
    Thomas Phinney Posts: 3,024
    Just in general, what you call “font filling” — which is to say, adding more glyphs in the same style as the pre-existing human-crafted glyphs — would seem to be one of the most obviously useful AI-driven features for type design, and something that existing sorts of AI tech seem well suited to.

    Presumably one could claim copyright on the human-generated aspects of the font, and disclaim it in other aspects. But many people might be reluctant to use a tool that could negatively impact the copyrightability of their work.
  • Ray Larabie
    Ray Larabie Posts: 1,459
    edited July 9
    Interpreting reports and generating scripts to fix bugs. Modernizing old fonts with marks, combining accent glyphs, and improved OT features. Modernizing vertical metrics, and family naming. I don't think you'd need to build new functionality into existing tools; it could make use of scripting features they already have. Currently Ai can interpret screenshots and bug reports, but it essentially holds your hand as you solve the problem yourself.

    And interpreting the FontLab 7/8 documentation: not a joke.
  • Opentype with AI to help achieve unique or complex features and AI as assistant alongside with Font Software to give some suggestions, like FontBakery+AI but live when working.
  • John Butler
    John Butler Posts: 322
    Identifying font pirates, finding their bank account information on the dark web, and transferring the appropriate license fee from them automatically without contacting them.
  • George Thomas
    George Thomas Posts: 656
    @John Butler — Nope. That would be illegal.
  • Thomas Phinney
    Thomas Phinney Posts: 3,024
    I am pretty sure that was intended as a dark joke, speaking to the frustration some type designers have with font piracy. “If only making pirates pay for the fonts they acquire was as easy as font piracy!”

    Or “What if type designers’ dealings with font pirates ignored the law — just like the pirates do?”
  • Nick Shinn
    Nick Shinn Posts: 2,284
    edited July 10
    Nobody “needs” AI or asked for it.
    Why embrace something so disastrous for humanity and the planet, for all the “productivity” it may produce?
    Perhaps we’ve gone too far
    Perhaps we should slow down
    And be where we really are
    —Richie Havens, Little by Little.

  • lcmchris said:

    […] an AI tool for type designers. It can currently generate simple fonts!  […]

    Apart from this surely being a nice toy for programing specialists, what shall that be good for? Considering the millions of cheap/free low-profile fonts already out there, there’s no need for that.
    I can’t see any case why AI should help me in designing or engineering fonts.
    What I would welcome is further developed built-in help in applications for the more technical tasks: creating diacrics and mark positioning definitions, OT code, and yes: language specific character sets – all those things which suck.
    Apart from that, what the font user community needs most is new keyboards. Created not with “AI” but by common sense and based on user needs … 
  • Mike Duggan
    Mike Duggan Posts: 241
    Although it would have been a good candidate, AI came along slightly too late for font hinting : ) .... as there is not much demand any more
  • Simon Cozens
    Simon Cozens Posts: 794
    Andreas Stötzner said:
    I can’t see any case why AI should help me in designing or engineering fonts.
    What I would welcome is further developed built-in help in applications for the more technical tasks: creating diacrics and mark positioning definitions, OT code, and yes: language specific character sets 
    It sounds like you would like extra algorithmic tools which can help you in designing and engineering fonts. Perhaps it should just not be called "AI", because AI is bad. Let's call them algorithmic tools.
  • Although it would have been a good candidate, AI came along slightly too late for font hinting : ) .... as there is not much demand any more
    Honest question: Do you think there is not much demand anymore because the use cases are few, the cost is high so people don't see it as a worthwhile cost, current auto-hinting does a good enough job, or something else?
  • Mike Duggan
    Mike Duggan Posts: 241
    Although it would have been a good candidate, AI came along slightly too late for font hinting : ) .... as there is not much demand any more
    Honest question: Do you think there is not much demand anymore because the use cases are few, the cost is high so people don't see it as a worthwhile cost, current auto-hinting does a good enough job, or something else?
    probably a combination of all of those things
  • One of the parts of creating a family where I struggle a bit and find tedious – even more than kerning – is when I'm generating a second master. For instance, when I'm creating a bolder version, it's very easy to end up deforming it to the point it significantly looses its character and that takes a lot of adjustment. Maybe this can fall in the font filling category.

    Another tool that I'd consider useful, would be something that checks thickness and weight consistency across the font. Or something that lets me link or define internal parameters for a font, so if I modify something, it automates that adjustment everywhere that becomes relevant for consistency.
  • Mark Simonson
    Mark Simonson Posts: 1,755
    edited 4:13AM
    One of the parts of creating a family where I struggle a bit and find tedious – even more than kerning – is when I'm generating a second master. For instance, when I'm creating a bolder version, it's very easy to end up deforming it to the point it significantly looses its character and that takes a lot of adjustment.
    In my opinion, this crosses over into design, which is the one thing I wouldn’t trust automation to do. What’s left to the type designer to do, if not design? For me, developing the weighting scheme of a type family is one of the most interesting and satisfying parts of type design.
  • Thomas Phinney
    Thomas Phinney Posts: 3,024
    I think some people would like to do a certain set of glyphs and have additional glyphs filled in by AI. Character set expansion, essentially.