arrowtype
commented
5 years ago Hey Chris, thanks for leaving an issue, thanks for the idea, and thanks for being thoughtful with your writing! It's definitely a cool thought, and one that is fun to consider as something that might be possible to add in the long run.
I know building a static font file is enough of a pain as it is, so I can't imagine the time it takes to build variable fonts, thus, feel free to tell me if this idea is an unreasonable amount of extra work.
This is the main thing stopping me from adding a serif variation to this family.
I'm not sure of how much you've drawn with bezier curves, but I'll try to explain in a way that might be understandable by others who come across this post, as well.
It takes a lot of time, effort, and acquired skill to draw a single font well. Letters are all created from bezier curves, which take some getting used to, and all letters in a font must be stylistically harmonious with all the others, and given the right proportions and margins to look good in text.
To make a variable font, there is a significant additional challenge: the drawings for every glyph must match in several ways, so that the drawings can be interpolated to create in-between drawings.
For a simplified example, if this font had only a single weight (ExtraBold), with the Slant axis, the two drawings required for the letter m would be like this:
(This is a screenshot from RoboFont, the software I have used to draw Recursive.)
Several things must match for interpolation to work well (or at all, often):
- the number of oncurve points must be the same (in this image, the blue/green/red points)
- the number of offcurve points must be the same (in this image, the yellow points)
- the contour order must be similar
- the "start points" must be similar
- the unicode values must match
- the number and names of "anchors" must match (these help accents like
dieresis, etc to attach)
However, to make the full five-axis Recursive VF, there are 24 compatible drawings for every glyph. So, that looks more like this:
(That is a screenshot from Prepolator, a useful RoboFont extension that makes it easier to match contour order and start point location.)
To add an alternate shape such as the m.italic with its "outstroke" tail, we have to repeat the process with 24 additional drawings:
This process varies somewhat for different letters. The uppercase and a few of the lowercase don't have any italic alternates. Any glyphs that look good with a "normal" width of 600 units it the Sans styles are only drawn 12 times, then copied over to the Sans masters:
Of course, that still requires some scripts to copy glyphs, and scripts to check that things haven't gotten out of order.
Other glyphs, like the i, require 24 compatible drawings (for the Sans shape), plus 24 compatible drawings for i.italic and and i.mono alternates, plus all of the diacritics based on i need alternates, too (though, luckily these diacritics can be built from scripting and components). So, all in all, under the hood the i must have 24 drawings for dotlessi, 24 for dotlessi.italic, and 24 for dotlessi.mono, and then scripts put these together with accents like the dotaccentcomb and all the rest. In one of the 24 masters, then, the i looks something like this:
Then, these alternates are connected to axis conditions via rules in the designspace. You can see the current ones here: https://github.com/arrowtype/recursive/blob/7fa1884493cb3b24709153b6058844083b8e47d8/src/masters/recursive-MONO_CASL_wght_slnt_ital--full_gsub.designspace#L28-L628.
All that is to say, if we wanted to add alternates like m.serif to make a slab-serif version of Recursive, we could do it in one of two ways:
- A smooth axis that required 24 additional sources (this would not be great for final filesize, but could be better if partial sources were used, via UFO layers – but this approach is harder to use with most drawing/checking tools)
- More realistically, there could be
.serifalts thrown into the mix, and a variable axis that is like the Italic axis, which would control whether these were activated or not.
It would probably require different alternate glyphs to be added for almost every uppercase and lowercase letter, all figures, and probably for quite a few symbols if we were being perfectionists about it.
I hope I've explained that clearly enough ... it has taken me time to wrap my head around all this, plus a healthy dose of good advice from people who are much smarter and more experienced than me. So, if it doesn't make sense, it's probably because it just takes some time working with these tools to grock it all.
With that said, this would be at least a few months of work, and add significantly to the overall complexity of the project. There are a few things that I am more keen on that a serif axis:
- adding support for cyrillic (also a huge effort)
- adding a width axis to allow condensed styles (also a huge effort)
- adding a blackletter variation (maybe a bit harder than serifs, but would be way more fun)
- starting my next font project and achieving some of these fun ideas with a design that makes it a little simpler to do, likely by leaving out axes like Casual and Slant
chriskirknielsen
demoy
Hi! Recursive looks like a dream font to me when thinking of performance on the web.
The sans and mono versions look great, but I feel like this font could be improved further with a serif setting. I obviously don't know how this could be achieved without a major rework, but thought I'd throw it out there. EDIT: As mentioned in this tweet, forking this to swap out the casual setting for a serif one might be an idea.
I feel this is already kind of there with the MONO setting set just above
0.5but it's not a true serif. If the mono version's serifs could be used without monospacing, that would be a great start.I know building a static font file is enough of a pain as it is, so I can't imagine the time it takes to build variable fonts, thus, feel free to tell me if this idea is an unreasonable amount of extra work.
Best, Chris