I tried a similar approach but via CSS for my "medieval content farm" (https://tidings.potato.horse), but for perf. reasons ended up only randomising the opacity and tint of its individual characters and removed the transforms.
My next step is to add those little padded areas hiding the empty spaces on the right side of the text[1][2], so the column of the text feels neater (horror vacui was a big thing then it seems).
Weirdly enough, this gets much easier with container queries and stable diffusion:
1. feed SD with some examples of patterns, then scale them to 10 or so sizes
2. then use those images as backgrounds for elements fitting the empty space in each line
3. select the right image using a container query
This should give them a neat but organic look. And hey I can brag about not using any JS!
If you haven't yet, you should play Pentiment. It's set in the early 1500s as the printing press is spreading across Europe, and each character's dialog text reflects how they would write (or print, for the people who operate presses). The text rendering is truly beautifully done, the ink dries a few moments after the text is written out and everything. I've been hoping for a technical blog on their text rendering techniques but haven't seen one yet.
Unrelated, and sorry in advance if this is unwanted advice... but I see on your homepage that you say your name is pronounced /'rafau pastoohshak/. But that's an English interpretation and /.../ is typically used to refer to the International Phonetic Alphabet. (/.../ is approximate and [...] is the precise version.)
I think the IPA one is /rafaw pastuʂak/ (note: /ʂ/ is very similar to /ʃ/).
The quote marks on that page are a real mess. Best practice is to use double curly quotes or single curly quotes—not backtick and apostrophe.
See https://www.cl.cam.ac.uk/~mgk25/ucs/quotes.html
("Summary: Please do not use the ASCII grave accent (0x60) as a left quotation mark together with the ASCII apostrophe (0x27) as the corresponding right quotation mark (as in `quote').")
I assume the author picked up the quotation mark habit from LaTeX. Backticks and apostrophes (single or double) get rendered as proper left and right quotation marks. It's also possible this was copy+pasted from a LaTeX source and they forgot to fix the quotation marks.
The italics on the page are also missing, for me at least. I am seeing them as obliques. The upright roman face is tilted, rather than using the separate set of italic glyphs. Some systems do this if the italic glyphs can't be found or something is misconfigured.
Funnily enough, I picked up the habit of using graves for my opening quotes from working with old Unix software like ELisp without any context. I think it's a fine practice, particularly when used as markup that's transformed into curly-quotes on the user-faceing side, and it's even seeped into my handwriting as exaggerated opening quotes. I'll still draw nice curly quotes or mirrored ticks when writing neatly, but for sloppy writing and markdown alike, I find the clear distinction between opening and closing delimiters to be more readable than using just single quotes. On this, reasonable minds may differ c:
> I find the clear distinction between opening and closing delimiters to be more readable than using just single quotes.
The correct characters to use are ‘/“ and ’/”. There is a clear distinction between opening and closing delimiters. I think you're falling for a false dilemma brought on by outdated software that is limited to ASCII (i.e., stuck in the 1960s).
That really stuck out to me too. Given the author's care with other things and the fact that the quote marks in the dissertation itself look fine, I assume it's just an unfortunate accidental error and not the result of ignorance.
It's not really an alternative. The appeal is the very organic and human look. The one you liked is certainly more apt for medieval stuff, but it still looks digital
Reading this really helps illustrate why and how a font is considered a computer program and thus eligible for copyright protection. I don’t think most laypeople (and even many programmers) intuitively think of it that way because they seem like a collection of little pictures, but so much work goes into these fonts. We are fortunate to have so many good ones to choose from today, and that someone has footed the bill for their construction, since most people never directly license a font.
> Reading this really helps illustrate why and how a font is considered a computer program and thus eligible for copyright protection.
Copyright protects all creative works, with very few exceptions. Whether or not a font is a computer program has absolutely nothing to do with whether it is eligible for copyright protection.
Even if the font was just "a collection of little pictures" (that is, a bitmap font), it would still fall under copyright protection, unless the pictures were all in the public domain already, or some other special circumstances apply (e.g. creative threshold not met, created by the government, etc.).
Interesting, I didn't know that some countries don't give protection to typefaces.
Does that mean that if you take a commercial font file and strip the bytecode, retaining only the glyph shapes, you can do as you wish with the result? And does that mean all bitmap fonts are actually public domain in the US?
Protecting a typeface itself would be super annoying. Every document you make, every billboard put up, every word on every sign, menu, signs on every vehicle on the road, would need to have its fonts explicitly licensed for those uses, the way webfonts are today. Complete with rules like “you can use our typeface on a sign, but if the sign is in a town over 10,000 population, we want $100, and if it’s a city over 1 million, we want $10,000. Oh, and all of this is per year.”
Also, good luck even making a free font without being accused of making a “derivative work” of every copyright-encumbered font that’s come before. (“look at those serifs!”)
I know I started the thread by saying I think fonts deserve copyright, and I do, but not the typefaces themselves, as that would drastically disrupt what you can create. It would be especially arbitrary considering most faces are derivative of what came before, stretching back centuries. To paraphrase a famous quote, “You didn’t build that, [Adobe]”
Stripping the bytecode may make the font look ugly at low display resolutions, though many modern font rendering programs use various auto-hinting algorithms to alleviate that (which were invented in order to circumvent the patents that could forbid the execution of the bytecode).
I don’t know what bytecode really is in this context, but a plain old vector image is a program too, so i don’t know of any process which would (in US law) let you convert a font into something public-domain.
(other than rasterizing text with it, at which point the output is not a “font” in the sense we usually mean when we use that term.)
Probably the majority of the software I use to write code is open source, but font authors will routinely ask for exorbitant amounts for a license, they often refuse to charge a flat fee, insisting that they get a higher cut if your app has a certain number of users. Just because an app is popular, doesn't mean it's raking in the dough to spend six figures on a font license. It's just bizarre that I can build an app that serves millions of users using tech stacks like Linux, HTML, JS, CSS, Rust, Go, with tens of thousands of open source libs and modules to choose from, none of which have per-user fees, the majority of which have no fees at all, of which millions of man-hours have gone into producing and maintaining, and a font could end up being the most expensive component of my app.
It's more comparable to illustration or photography, or even to a written body of html or js, than it is to the programming language. It's the output of an applied specialist technician for a specific purpose.
So similar to those other things you can find a free or cheap generic off-the-shelf one that suits your use, open source fonts exist. Or you can pay for a specific one that suits it better.
Font design is a domain several times older than the oldest reasonable precursor to computers. It's imo shitty to stomp in and demand it conform to our idea of how licensing should work.
> It's more comparable to illustration or photography, or even to a written body of html or js, than it is to the programming languag
Yeah again I can download about a few hundred thousand open source tools like ffmpeg, git, Linux, Chromium, web servers, databases, queues, web libs, etc and almost always the best stuff is open source. Almost never is there some kind of per user fee just for copying code, and if it is, almost nobody is using it.
It's not shitty for me to call out greedy excessive licensing practices.
Font forge is open source, which is a tool more comparable to those than a completed font is. You're not entitled the results of the work of other specialists, even if you've become accustomed to that in other areas. And again, do you give away your completed work? Does no one pay you for the expertise you use for them?
A font is much more similar to a completed web app than it is to the tools used to create that app. You don't expect someone to make an app for you for free, even if there are apps available for your use for free. Same thing here.
Typographers aren't getting rich lmao. They're trying to pay the rent with their skills same as everyone.
And even proprietary software used in programming web servers rarely ever attempts to charge a "per user" fee or six figures for a mildly popular app to use.
In case this helps anyone who’s been in this boat: I found that Monotype, who owns a LOT of fonts, offers far better terms for web and app licenses on fonts dot com, compared to myfonts dot com, which they also own. I licensed a bunch a couple years back that would have cost a fortune in recurring license fees at myfonts, which has great SEO so it’s always the first place you find.
One-time purchase from fonts dot com. Yes, it was still scaled based on monthly impressions, but at least you could pay and use it forever.
The other thing I found was that, on the corporate scale, Adobe Fonts (aka Typekit) is a bargain. Basically an unlimited web and app license for everything.
My first job, between high school and university, was at a bus company, helping with scheduling and timetabling, in the mid 90s. Me, teenage geek... with a Sparc 5 on my desk.
What I wish for is rather than encoding the wavy imperfections in the font, they should be procedurally generated at rasterization time. This way, the graininess can stay at the same resolution regardless of the font-size. As it is now, if you have a 120pt font, the rough edges are magnified.
Is there a way to specify something like that in the font language? Or does the rasterization engine have to implement that?
I was already writing a comment about procedural generation before I saw yours. I was thinking about placing shapes on the boundary of a basic character shape, and placing random points within those shapes which then have lines/splines drawn between them. You could change the size of the shape depending on how much you want the shape to vary. That way you can still exercise control over what specific features are preserved. I'm sure someone can come up with something better.
Anything procedurally generated is antithetical to professional font design. Fonts at different sizes are carefully fine-tuned exactly for the reasons that you mentioned, i.e., small features at larger sizes start looking wrong.
In this case, the digitized font is taken not from the metal types themselves, but from specimens printed on paper. The paper was less uniform and smooth than modern paper, and there was a noticable amount of ink bleed. Since OP is interested in the "character" given to the letterforms by this era of printing technology, it makes sense to suggest some way of simulating this uneven bleed and spread.
Usually, yes, but this font tries to reproduce the imperfections of text printed centuries ago (or, to quote the article, "look badass with its kinks and nicks"), so having two glyphs of the same type look exactly identical feels a bit jarring. Same as with other "rough" fonts which try to emulate stencils, handwriting etc. All of these would benefit from some randomness...
There used to be. In ‘Type 3’ fonts, each glyph is defined by a Postscript routine. They were not supported by any of the major display systems (Windows, Mac, X). Maybe NeWS or NeXT did; I don't know.
Yeah, I was idly wondering if it might be fun to try rendering fonts through a basic physics simulation of inks and paper types. It would probably be wasteful to do most of the time in font rendering, but it might be fun to try when doing things like rendering final slides or PDFs (that are themselves not intended to be sent to printers, because that would be silly).
This is fantastic work. It's very hard to find historic fonts that haven't been overly modernized. For fun, I like to restore and re-print old playing card/tarot decks and other ephemera for use by a modern audience. In order to make them more legible it makes sense to use translated or reconstructed text. Having fonts like these readily available makes this work so much easier.
This font is slightly too messy for me, but I agree that "modern typography" is often too soulless. I'm also hoping for a renaissance of serif fonts - I'm not really buying the narrative that they're not suitable for the web.
Not a big typography expert, but one font I like and try to use often, is Baskerville.
On the other hand, while I love LaTeX, I absolutely abhor Computer Modern. Maybe it's not the font's fault, but rather the fact that by using it every paper ends up looking the same - but it just bores me to death.
> I'm not really buying the narrative that they're not suitable for the web.
Good, because from a technical perspective, you don't need to. The legibility argument has been less and less relevant as screen resolutions have gotten good enough to display serifs in a reasonably print-like manner. (Which was arguably around 2010–2014 when Apple normalized “Retina” displays above 200 dpi, but is especially true today, now that the majority of web browsing occurs on mobile devices).
What makes the difference is of course the cultural and branding associations with serif typefaces. But actually thinking about it now, I feel like I've been seeing a lot more serif typography on the web recently, even in modern, tech-related contexts. And looking at the big example of theverge.com with its quirky redesign, I think you might just get your renaissance soon :)
Likely because while Computer Modern was a great font for 1978, its last update was in 92 & it’s painfully outdated & low quality compared to modern digital typefaces.
Its continued usage is a crime given how good Latex’s typesetting is.
Frankly, I’m not a huge fan of its stale "Didone" style either. It’s a style that does better in advertising or magazine covers than for reading.
Computer Modern is okay on paper, but I find it very frustrating to read on screen. I wonder if the Fell types are similar ‐‐ the scan of the thesis, where the letters aren't fit to the pixel grid, looks a lot nicer.
I do like Baskerville, but I am a little sad that the only open source version I know of, Libre Baskerville, changes the design so much for the screen. Caslon is in the same boat.
I think the Fell types actually read pretty well on screen. (In contrast to Computer Modern, a typeface I really like, but agree does not read as well on screen.)
I actually found this really pleasant to read. My eye seemed to be able to move from each word to the next more smoothly. It is interesting to note how the slightly uneven weights on each character are similar to fonts designed to aid Dyslexic readers like OpenDyslexic. I wonder if that is why. https://en.m.wikipedia.org/wiki/OpenDyslexic
There are probably other factors (some shapes just are more readable...) but I believe it's a big part of it yes.
There was a typography trend in the late 90s, 2000s, and you still see echoes of it, towards "mathematical," "geometric" design in fonts. Eg using exactly the same curves, exactly the same bowl size and proportion, exactly the same serif etc across all characters. Now this is mostly regarded as a dead end. The text ends up feeling inhuman or bland, and the readability gains are at best negligible but often simply worse. It seems that the small inconsistencies aid quick character identification, if not too distracting.
Right now the trend is more towards specific, intentionally introduced inconsistency. Most of them are too subtle to notice unless you're specifically studying the font, but ideally in aggregate they give it a certain organic feel and differentiate similar shapes used in different characters.
I took slight offence at the passage "a live performance of a great symphony, in which musical instruments played by humans are never perfectly in tune and occasionally quite far out of tune" after mentioning the Wiener Philharmoniker. I don't think they ever play out of tune.
I did hear orchestras play out of tune occasionally, but it has never been an agreeable experience. It never made me think "how human". Instead, hearing a botched entry in the chorale in Mendelssohn 5 helped me understand the brass players' addiction to beta blockers.
In LaTeX, the open quote is denoted by a grave accent and the close quote is denoted by the apostrophe. Look back at the article and you'll find that the author consistently uses these characters in this manner. Why they didn't get processed to the proper Unicode characters, however, is beyond me.
A typeface is not a copyright protected work in the United States of America at the time that I wrote this [0]. However, files that font rendering programs can use to display typeface glyphs are protected works [1].
[0] https://www.copyright.gov/circs/circ33.pdf "Works Not Protected
by Copyright" page 3, section 2 ; though, it does call out limited protections for heavily ornamented calligraphy and the like
This font's ragged looking nature — is it primarily because of the way the font was printed at the time? I would assume there is some loss of clarity and other artefacts introduced by how metal type printing prints. It seems like these fonts were created only by looking at printed works. Could the original metal blocks (sorts/matrixes/punches) still be around?
Sure enough, if you go from the end of the blog post straight to the thesis, you can totally hear the introduction of Also Sprach Zarathustra coming out of the title page.
Reading this page I actually wonder, why don't we do it all the time. It never occurred to me before, and I'm even used to the idea that bland sans-serif fonts are somehow required of the screen, but web-fonts look like shit compared to your typical older book typeface. This page is simply more pleasant to read.
> that one should never change the letterspacing of the lowercase letters
But with the microtype package it is acceptable to slightly expand or contract the letters to avoid hyphenation. Typically the limit is set to around 0.025em.
Anecdotally if you are applying letterspacing to uppercase letters, 0.025em is also a good default value to use.
- "It is certainly not as far-fetched as Leonard Bernstein's controversially slow reading of Nimrod from Elgar's Enigma Variations with the BBC Symphony Orchestra."
That link got Google-holed, but I think it's the same content as this one (?)
> The very first specimen book of the Fell types, printed at Oxford University Press in 1693, when J. S. Bach was only 8 years old. Only four copies are known to exist. Furthermore, many of the original punches and matrices, some of which were made of wood, have been lost (Oxford University Press 1900), so these specimens are the only record.
Ah so this is THE Dr Fell from an old remembered nursery rhyme
I'm usually a fan of Windows-style font rendering that snaps on-screen text to the pixel grid. But for this typeface it seems to heavily exaggerate the roughness as the different serifs snap all over. The pictures of the print pages look much nicer.
Something about taking a screenshot on a scaled display makes my first image (at 125% scaling) look bigger than it should be; at 80% magnification it looks about the same size as it does in my browser, so edges may not come across quite right. (100/125 = 0.8, so that’s probably not a coincidence.)
I set scaling to 100% for the second image, and that one looks true to life size, so it’s probably the better reference.
Edit: On comparison, it’s not the serifs. Maybe it’s the lighter overall color on the page, or maybe it’s the way that autohinting exaggerates the contrast between thick and thin strokes.
This is exactly what I miss about modern computer-typeset sheet music. There's a lovely organicness to the very slightly splotchy noteheads etc. in old sheet music, and modern scores feel very antiseptic to me by comparison.
I’ve used Glyphs before for drawing type and can’t recommend it enough. Reasonably priced, quite powerful, fast and fluid to use. Drawing—even re-drawing—type is almost addicting. When you get the Bézier curves dialed in just right it’s so satisfying!
glyphtracer [1] can help with this, but letter shapes are only a small part of what makes a typeface. (Typography is about whitespace, or as Miles Davis said, “it‘s not about the notes you play, it’s the notes you don’t play that matter”.)
Like the lack of typeface requirements, I think there was also a belief that the university regulations still allowed phd theses to be submitted in Latin, though I’m not aware of anyone trying it.
Are those statements about “eye strain” for serif, sans and sans serif fonts scientific? Every time I see them they are not supplied by evidence but by some mumbo jumbo.
The long-standing debate over serif versus sans-serif for readability is likely less important than we thought. Other factors, like the height of the letter or spacing between them, change how easy a text is to read more than the presence or absence of serifs.
Yes and no. To a certain extent there is brain familiarity with letter shape, but screen resolution plays a role. Sans fonts are simply more legible on low-resolution displays.
I balanced a consulting job and my PhD, so thesis time was incredibly productive. When I didn’t want to do one, I’d become very productive on the other, and when I got tired of that, I’d switch and get very productive again.
> ``. . . anyone can do any amount of work, provided it isn't the work he is supposed to be doing at that moment." -- Robert Benchley, in Chips off the Old Benchley, 1949
Thinking about it, i suppose what i actually did to avoid working on my thesis (and pay the bills) was get a part-time programming job, and then that job became full-time, and i never finished the thesis. So, in a sense, fifteen years later, i am still procrastinating.
For some reason the last “x4” confused me but then I remembered the z-axis. Then it had me thinking how a 4x4x3 Rubik’s Rectangular Cuboid would work. It can’t be turned 90 degrees on one axis. Anyway, for the next time you have a thesis to avoid, maybe.
To be fair, I wasn't just procrastinating from writing; I was also dealing with being diagnosed with a life-threatening illness. I semi-officially took a retroactive leave of absence for a term (I was a bit behind schedule, mentioned the medical issue as having contributed to delays, and the head of the department said "we could say that you had a leave of absence but I don't want to bother with the paperwork if you don't want to").
My next step is to add those little padded areas hiding the empty spaces on the right side of the text[1][2], so the column of the text feels neater (horror vacui was a big thing then it seems).
Weirdly enough, this gets much easier with container queries and stable diffusion:
1. feed SD with some examples of patterns, then scale them to 10 or so sizes
2. then use those images as backgrounds for elements fitting the empty space in each line
3. select the right image using a container query
This should give them a neat but organic look. And hey I can brag about not using any JS!
[1] https://i.pinimg.com/originals/2f/46/2d/2f462df256f4bd509b48... [2] https://i.pinimg.com/originals/ac/68/a8/ac68a86da5ec457960d0...