Strategies for drawing a variable sans with weight, width, and preferably slant axes

Ori Ben-Dor

Hi,

I'd like to draw a variable sans-serif typeface with weight, width, and preferably slant axes.
I ran a small-scale pilot, and two main issues came up.

But before I go into details, if you can think of relevant open-source projects that I could dissect, please let me know.

First issue:

Ignoring the slant axis for a moment, I want my narrow ovals—but not my regular-width ones—to have a middle straight segment (see the figure). So the transition from wide to regular-width to narrow isn't linear. Hence just 3 or 4 masters on the corners of the 2-dimensional weight-width design space (corners = extremes on both axes, e.g., light narrow) are insufficient. Adding a master in the middle (regular weight, regular width) didn't help. So I tried masters in the middle of the edges (an extreme on just one axis, e.g., regular-weight narrow), but then I felt I didn't have enough control over the interpolated corners and it became hard to get them right. I'm reluctant to draw both corners and edge middles... Any suggestions?



Second issue:

Bringing the slant axis back, I don't see how kinks can be avoided. Consider for instance upright regular-width- vs. narrow n; the narrow one calls for a higher shoulder, which means the straight segment of the leg and the tangent handle that emerges from it don't share the same length ratio across the width axis (see the figure). But then, when I slant them, those different ratios mean a kink for interpolated instances. Are there any solutions or workarounds?

Alex Visi

That setup is hell on earth but possible. 

I doubt you’ll get nice forms across the whole alphabet without normal width masters for each weight/slant (would love to see if you’ll find a trick to get away without it, but...). Actually, good chances some letters will end up having a lot more masters and /O will be your smallest concern.

Also, what are you going to do with slanting those extra points in the middle of ovals — if you hide them in the same positions, you may get ugly hinting, and if you leave a small gap, you may need to find how that gap will be slanted.

For the n, did you get kinks in practice? It should not be a problem if you keep the ratio between upright and slanted of the same width/weight the same, there should be no kink. Because you don’t change ratio and angle at once.

Drawcard

For the first issue, take a look at Recursive:

https://v-fonts.com/fonts/recursive (the 'casual' axis reshapes the /O from a straight sided version to a rounded shape). Here are the source files for it: https://github.com/arrowtype/recursive

These are commercial licenced examples but also worth noting:

https://v-fonts.com/fonts/loos-variable 

https://v-fonts.com/fonts/antarctica-var
https://v-fonts.com/fonts/extenda-variable

https://v-fonts.com/fonts/widescreen-vf

Thomas Phinney

Alex Visi said:

For the n, did you get kinks in practice? It should not be a problem if you keep the ratio between upright and slanted of the same width/weight the same, there should be no kink. Because you don’t change ratio and angle at once.

Ah, clever. I briefly thought this was a mistake, but then I realized what Alex was saying.

Ori Ben-Dor

@Drawcard, Thanks for your suggestions!

@Alex Visi, Good question about those extra points in the middle ovals, I haven't thought it through (I didn't test hinting in my experiments).

@Alex Visi, @Thomas Phinney, Right, but that means I need a slanted master for every single upright master, doesn't it?

@Thomas Phinney, Science Gothic covers the same design space, doesn't it? Have you run into similar issues? How have you dealt with them?

Thomas Phinney

Ori Ben-Dor said:

@Alex Visi, @Thomas Phinney, Right, but that means I need a slanted master for every single upright master, doesn't it?

@Thomas Phinney, Science Gothic covers the same design space, doesn't it? Have you run into similar issues? How have you dealt with them?

1) Yes, you do.

2) Yes, it does! It has a slanted master matching each upright master, with no change in height-for-points on those slanted masters. In a few cases this means that there have to be overlapping contours or other complications.

I often wonder if Science Gothic has way too many masters. Yes, the design space is so large that intermediate masters became well-nigh unavoidable. But… the font is really big as a result. Not a great thing in a web font.

Having a slant axis instead of a separate oblique/italic font is also a tough choice. It reduces total file size a bit. But it also means that, absent sophisticated subsetting, one can’t reduce file size by skipping the oblique/italic.

At least you don’t have a contrast axis!  

Ori Ben-Dor

1) 

2) Thank you for this answer, I'm going to download Science Gothic's source and study your design space.