Renaissance Font Reproduction for Historical Music Notation Software

Sam Gossner

Hello!

I run a small music software company; one of our recent projects is a web app designed to facilitate the retypesetting of 16th century music notation, from the dawn of musical typesetting to its fall from grace. We are looking for a freelancer to assist in creating accurate digital reproductions of period text fonts for use within our app.

Background

Between about 1500 and 1650, music was normally typeset using moveable type; subsequent changes in the needs of notation led to engraving becoming the standard (through today), though there are examples as late as 1900 of moveable type used to print music. There is a modest community of scholars, enthusiasts, and musicians who are dedicated to performing and teaching this old typeset music and its unique notation style, but the sad state of many scans and facsimiles makes their task very difficult.

Our goal is to allow users to take such damaged or flawed scans of original books and re-typeset them in a clean but still historically-flavored appearance, with the specific look and spacing of typeset music and glyphs matching those of the original publisher.

An example of our application's current output (with text added after):

The Problem

While I've had reasonable success creating the music glyphs as shown above (there are currently five 'fonts' of such, covering different publishers in different regions and eras), we've had trouble sourcing [open source] text fonts which are entirely appropriate for the period (i.e. without excessive modern aesthetic tweaks) and have all of the unusual/deprecated characters necessary to create an accurate facsimile. As a result, I've been working from my own reference photographic scans and public domain scans of period texts to create historical fonts that fit the period and locales accurately, and include extra characters and ligatures dropped in most modern recreations.

The problem is, I'm not a typeface designer; I'm not even a graphic designer period— I'm a musician/audio-engineer with just enough of an eye to see that the results are okay but could be significantly better, but not enough to fix them properly. Kerning and spacing is always a bit of a mess, sizes and weights remain troublesome, and OpenType features remain mostly a magical mystery to me... hence why I'm here.

The Job

I'm looking to hire a freelance individual or group of individuals to assist in making these fonts consistent and more professional and synthesizing adequate characters from scratch where not available. Experience and interest in early printing practices and pre-18th century typeface design language is desirable. This is a remote, contractual/freelance job.

Unfortunately the end product is very niche, and despite a trickle of support from a handful of generous end users, I'm more or less supporting the entire project personally out of pocket. I certainly intend to pay for any services rendered, but our budget is limited is all. I would alternately be open to letting whoever helps commercialize the fonts on their own terms, though I recognize what we're doing here is probably too niche to be all that salable.

If you have any questions or are interested in helping, please reach out to me at sgossner@versilstudios.net, direct message, or below. Thank you for your time and consideration!

Here are a few of the fonts I've put together so far, in their current rough WIP state.

Jasper de Waard

The fonts you've created don't look half as bad as I'd expected

This sounds like a really fun and interesting project. I don't have the time to help you, but I really hope you find someone skilled to help you out. If that doesn't work out for whatever reason, you could consider posting one or two of your fonts here on typedrawers for feedback and work on them yourself. Personally, I quite like nr 1 and nr 4.

Also, I wouldn't be surprised if the kind of type your looking for in some cases does already exist. You may simply not have found it yet. Again, I can't really help you with that, but I'm sure some of the knowledgeable people here on typedrawers can.

Anyway, good luck!

Oliver Weiss (Walden Font Co.)

At that time, kerning and spacing in utilitarian printing weren't all that great. It looks like you captured the general look and feel. I think your fonts are pretty good as they are! 

Sam Gossner

Jasper de Waard said:

The fonts you've created don't look half as bad as I'd expected

This sounds like a really fun and interesting project. I don't have the time to help you, but I really hope you find someone skilled to help you out. If that doesn't work out for whatever reason, you could consider posting one or two of your fonts here on typedrawers for feedback and work on them yourself. Personally, I quite like nr 1 and nr 4.
... ...
Anyway, good luck!

Oliver Weiss (Walden Font Co.) said:

At that time, kerning and spacing in utilitarian printing weren't all that great. It looks like you captured the general look and feel. I think your fonts are pretty good as they are! 

Thank you both, I appreciate that a lot!

The main issue I'm really running into is when there are missing characters; I have no experience actually designing a glyph from scratch so it's been pretty rough when something is missing. Likewise it takes many hours of tweaking for me to try to fix issues with character sizing or weight, and even then I come back days later and feel compelled to change things again. Someone with experience could probably do things 'right' a lot faster, while also giving me more time to work on the areas I am more competent in.

One person has reached out so far, so hopefully things will work out soon.

André G. Isaak

As Jasper pointed out, there may be cases where suitable type already exists.

You might want to have a look at Gille Le Corre’s fonts, which cover a range of historical periods.

https://creativemarket.com/GLC

André

Sam Gossner

André G. Isaak said:

As Jasper pointed out, there may be cases where suitable type already exists.

You might want to have a look at Gille Le Corre’s fonts, which cover a range of historical periods.

https://creativemarket.com/GLC

André

Thank you for the suggestion, André! Those look very nice indeed and many would fit the period, but unfortunately the licensing terms we would need are much broader than I assume most commercial foundries would be willing to agree to, at least at a reasonable price.

Basically we'd be embedding the font directly in a specialized word processing web app, with unlimited users and unlimited time span, with users able to freely export PDF's including the font embedded, including commercial sheet music publishers who might use it in a book of music and so on. I think the licensing cost would be far greater than what we could afford, or otherwise likely an insult to the designer for their immense time invested in creating their fonts if we could afford it.

Plus, there are some specialized symbols and marks only found in musical lyrics typesetting, which I do not see in these fonts. We would have to commission the creator to add them, or at least get permission to hack them in ourselves.

It is a 'catch-22': we are making a very obscure product only useful to maybe a few hundred or thousand people in the entire world and are unlikely to see much profit, but what we need to make that product requires highly-specific, historically-informed assets with an extremely permissive license. Hence we have only examined FOSS or pseudo-public domain options.

That's why I figured making the bulk of the font assets myself, then have someone more experienced step in and polish them a little, would be the most reasonable solution.

Thanks again for your help!

Jasper de Waard

If licensing is your issue, you could always email designers/foundries directly. Simply explain your situation and ask for a price. Standard licenses are often written relatively conservatively, so that more liberal exceptions can be made in specific circumstances. I suspect some type designers will also be enthusiastic about your cause, and therefore willing to bend their own rules a bit  

Scott-Martin Kosofsky

What you are facing here, Sam, is something that all designers of historical revivals come up against: determining the level of signal-to-noise ratio you are trying to preserve or move away from. Will it be a pimples-and-all reproduction, or an interpretation that enables users to deploy the fonts for a current purpose, which in this case would be as an aid to performers of historical music? 

What you’ve done (and I think it’s an excellent start) might fall too far into the pimples-and-all category for some tastes, including my own. Where that becomes most evident is not in the note heads or stems, but in the staff lines, which repeat the same distortion again and again, the slight slant and swell toward the lower right, likely the result of copy and paste—an action that is an artifact of modern technology, not at all historical.

It’s my feeling that every successful historical revival is an act of interpretation, not mechanical reproduction. In judging old type that was not “perfectly” printed, one tries to develop a personal sense of its essence, asking ‘what was it that would have been judged as ideal in its own time?” And, of course, you’re free to move on from there, especially if you are trying to adapt the type to modern usages.

Ross Duffin, a musicologist who teaches at Case Western Reserve University, published a list of digital fonts, each of which he instigated, based on early printed music: https://casfaculty.case.edu/ross-duffin/homepage/fonts-for-early-music. It appears that all of these types are encoded to work in music composition software such as Finale, which, I believe, generates its own staff lines.

As someone who grew up in the field of Early Music (I was one of the founders of the Boston Early Music Festival), I realized long ago that its essence was to get at the expressive truth of the music itself, which is greatly aided by serious inquiry into historical practices. I found that the "costumey" aspects of the field were distractions that were best avoided.

John Hudson

There are two separate things to consider here:

1. The musical notation system as a system, which includes vocal text when present; independent of a particular medium.

2. The way in which that notation system was printed in the 16th and 17th Century, and hence what it looked like in that medium.

In the original post, Sam writes:

Our goal is to allow users to take such damaged or flawed scans of original books and re-typeset them in a clean but still historically-flavored appearance, with the specific look and spacing of typeset music and glyphs matching those of the original publisher.

So that is clearly focused on (2). And it is fine for that to be the goal, but I do find myself wondering why that is the goal? Having the stave lines break, for example, is interesting to me, as a type maker, in how it exposes the composition method of the historical typecasting and setting. But in terms of engaging with the musical content I find it mostly distracting noise.

Sam Gossner

Scott-Martin Kosofsky said:

What you are facing here, Sam, is something that all designers of historical revivals come up against: determining the level of signal-to-noise ratio you are trying to preserve or move away from. Will it be a pimples-and-all reproduction, or an interpretation that enables users to deploy the fonts for a current purpose, which in this case would be as an aid to performers of historical music? 

What you’ve done (and I think it’s an excellent start) might fall too far into the pimples-and-all category for some tastes, including my own. Where that becomes most evident is not in the note heads or stems, but in the staff lines, which repeat the same distortion again and again, the slight slant and swell toward the lower right, likely the result of copy and paste—an action that is an artifact of modern technology, not at all historical.

It’s my feeling that every successful historical revival is an act of interpretation, not mechanical reproduction. In judging old type that was not “perfectly” printed, one tries to develop a personal sense of its essence, asking ‘what was it that would have been judged as ideal in its own time?” And, of course, you’re free to move on from there, especially if you are trying to adapt the type to modern usages.

Ross Duffin, a musicologist who teaches at Case Western Reserve University, published a list of digital fonts, each of which he instigated, based on early printed music: https://casfaculty.case.edu/ross-duffin/homepage/fonts-for-early-music. It appears that all of these types are encoded to work in music composition software such as Finale, which, I believe, generates its own staff lines.

As someone who grew up in the field of Early Music (I was one of the founders of the Boston Early Music Festival), I realized long ago that its essence was to get at the expressive truth of the music itself, which is greatly aided by serious inquiry into historical practices. I found that the "costumey" aspects of the field were distractions that were best avoided.

Hi Scott,

Thank you for taking the time to share this feedback! What you say is certainly very true and a key factor in my work on this project; much of what we have done is indeed a compromise between authenticity and ah, legibility.

For some background, I've been a part of the Boston early music scene for about a decade now (sackbut player), though my main work is in sound recording and sample library development (somewhat related field to font design). I also work as a printer for Von Huene and a few other clients and have spent a lot of time studying both facsimiles and the various modern compromises attempting to re-typeset this music. I am of course eminently familiar with the excellent work of Ross Duffin, who himself contributed to paving the way for modern typesetting of early notation with SCORE and the Ogni Sorte editions in the late 70's/early 80's.

It is my opinion that there is no reason to mix between modern notation and early notation- that it is perfectly fine that some people or groups wish to read fully-modernized editions, while others prefer to read facsimiles; but trying something purely in the middle like modern symbols and spacing methods without barlines in my honest opinion takes the worst aspects of both and puts them together. I read both forms of notation fluently and regularly, and find they both provide certain advantages in cases, but something which is too 'middle-ground' tends to read in an uncomfortable and disconcerting way.

I believe there is plenty of room for people interested in performing only from facsimiles, who wish to perform a work from a completely unreadable facsimile/scan, to be able to retypeset it in such a way that does not suck out the entire original spirit of the typesetting. There are innumerable works which do not have modern transcriptions OR clean, readable facsimiles which languish unread on IMSLP and in libraries. Using modern notation software to create a facsimile is and always has been an uphill battle in this regard, hence efforts like EMS Serenissima by Elam Rotem, and modern editions are only good for those who wish to use them, with their many compromises and tweaks. ENT is also an option for teaching the basics of early notation, something which is made much more difficult because of the poor state of many facsimiles and scans.

The real thing which sets ENT apart from notation software is that it is not too smart for its own good. It does not consider 'measures' or even know what they are. It only sets symbols in the order you tell it to, and has some automation to handle spacing and custos placement at the end of lines. That is why it is a typesetter, not an engraver or 'notation program' like Finale or Sibelius; it plays by a completely different rule book and gives a different result which is much closer to what would come off of a press in 1550, rather than an engraving from the 1950's.

Regarding the staff lines, those are procedural and randomized (each line segment is its own graphic), and a completely separate entity from the actual note glyph. Staff lines can be selected from four variants so far, and we are planning to add options for thickness. In this way, the staff lines can be as smooth or as rough, as regular or as ragged, as the user wishes. We can add an infinite number of variants and types of staff line shapes, and the user can change the spacing between the lines to make them completely continuous or widely spaced. Here are four of the variants at identical spacing, but you could if you wish make the lines completely smooth without any distinction between the pieces by reducing that minimum spacing value (to appear like triple-process printing of Petrucci, where lines were a separate element):





If you study the manuscripts in detail, you will find the exact same swell behavior in many books, this one shows the 'scallop' goes regularly down on the top and up on the bottom lines, a tweak we hope to implement for this particular staff line style soon. This feature is so prominent in many works it seems Elam Rotem implemented a very similar taper in EMS Serenissima perhaps for this same reason:


I have spent about a year making detailed measurements of the ratios and character of each notation style. There are five in all currently, covering the publications of Phalese the Younger, Rhau, Tini, Ledertz, and even a 19th century Ditson and more on the way. My goal is to provide a sufficient number of musical fonts (12+) to the extent that a user can more or less select an exact publisher or city/century and work from their type, like stepping into their print shop.

I should also mention ENT can flawlessly handle N-length ligatures (without 'hacks' of any kind), something I don't believe any modern notation program can do. It even slants at identical angles to ensure long slanted forms remain readable, and has no problem with half-coloration or other factors.

The music 'fonts' aren't actually fonts, but banks of SVG's. In this way, we can quickly implement all sorts of special cases (such as a custos which automatically switches to shorter versions as it is higher in the staff, to match the original) and randomized variants if needed, as well as reducing the amount of editing needed to tweak positions or behaviors.

There is a fully-usable demo (with the single font 'Tini' available) on our website if you are interested in trying any of these features:
https://typesetter.earlynotation.com/index.html

Anyway, I digress. If you have any more questions about the typesetter itself, please feel free to reach out, I am always happy to discuss our process and my work via e-mail. This thread is more intended to discuss the accompanying text fonts, not the notation. Thank you again for sharing your thoughts!

Sam Gossner

Jasper de Waard said:

If licensing is your issue, you could always email designers/foundries directly. Simply explain your situation and ask for a price. Standard licenses are often written relatively conservatively, so that more liberal exceptions can be made in specific circumstances. I suspect some type designers will also be enthusiastic about your cause, and therefore willing to bend their own rules a bit  

Thank you Jasper, I will consider reaching out to some designers based on your advice, though as someone who creates similar products (musical sample libraries), I would feel badly asking them to agree to some much lower sum is all.

Sam Gossner

John Hudson said:

There are two separate things to consider here:

1. The musical notation system as a system, which includes vocal text when present; independent of a particular medium.

2. The way in which that notation system was printed in the 16th and 17th Century, and hence what it looked like in that medium.

In the original post, Sam writes:

Our goal is to allow users to take such damaged or flawed scans of original books and re-typeset them in a clean but still historically-flavored appearance, with the specific look and spacing of typeset music and glyphs matching those of the original publisher.

So that is clearly focused on (2). And it is fine for that to be the goal, but I do find myself wondering why that is the goal? Having the stave lines break, for example, is interesting to me, as a type maker, in how it exposes the composition method of the historical typecasting and setting. But in terms of engaging with the musical content I find it mostly distracting noise.

Thanks for sharing these thoughts, John.

As I mentioned in my answer to Scott, the staff lines are fully configurable by the user (minimum spacing between type pieces can be configured in 0.001" increments or equivalent mm), including eliminating the spacing or making it more severe. Personally I find a very small space, about 0.002-0.004" minimum, actually helps with the legibility for some reason, at least as someone moderately familiar with typeset notation (I should caution I am NOT an expert in the field, just a very passionate hobbyist). Historically the variance I've observed so far is as great as 0.02" and as little as 0.001", I believe mostly a factor of the age/wear of the type.

[[I should note, the depicted form of typeset notation with spaces between lines is a 'single-process' form, where the note itself and staff lines are a unified piece of type; there was also a multi-step process was used generally earlier, in particular by Petrucci, wherein staff lines, notes, and text were each a separate process; pretty remarkable stuff for 1500!]]

The primary difference from most existing solutions is that ENT 'thinks' and 'acts' like a typesetter, not an engraver or a notation program. It has no pre-conceived notions about what it is typesetting, it is just following the order of symbols you give it, then it spaces them out in a logical order.

The problem is that while you can represent historical notation in modern notation analogs, there are some aspects which cannot be translated, such as the problem of the 'measure', a metrical invention conceived late in this period wherein consistent groups of beats are segregated with a vertical line. Before the use of measures and lines dividing them, notes could be of any length, afterwards, they must be 'cut up' so they fit into these measures. The result, notes 'tied over barlines', is considered by some to be ugly and a contradiction of the way people thought about notation. So, a wide range of attempts have been made over the past few centuries to 'adapt' modern notation to handle early notation, most of which in my honest opinion seem to be worse than either-or.

Anyway, once again, I digress- this topic should remain focused on the text fonts and questions regarding those. I welcome any questions about the musical notation itself via e-mail, or if there is enough interest, I could create a new topic in a more appropriate part of this forum for further public discussion if that is of interest.

Andreas Stötzner

There seem at least two sides of the challenge you venture for (which I find admirable). A.) the mere optical aspect of reproducing the appearance of the old printed sources; B.) the underlying system of the characters used for this kind of music composing, which has been based on a special lettercase’s content in the old days, but this system has no equivalent so far in our modern text reproduction architecture, which consists of character encoding, composing engine and suitable fonts. Which leads inevitably to the question: on which sort of character encoding your development is based. I presume it is a proprietary hack solution (similar like Sibelius or Finale deploy, to the day), please correct me if I’m wrong. That would enable to reproduce the optical surface of the sheet music in question, but not to convert the semantic content of those prints into actual transferable text information in the sense, in which we deal with all sorts of other texts, based on a standard character encoding.

It has been a failure to day to develop a viable composing system for sheet music based on systematic character encoding, mainly due to the fact that (western) musical notation is apparently too complex, similar as ancient Egyptian or Mathematical notation, which also embody a multi-dimensional text structure opposite to our common language-script-notations which are (more or less) mono-dimensional linear text-string systems.

Having said that, I’d like to point out if only one of the manyfold musical notation systems out there would be suitable for transmission into our modern technology, it would certainly be the lead-type composed music of (mainly) the Renaissance era. Because here we have a definite limited set of descret characters – and that is the main condition for defining a standard character encoding scheme on which fonts can be built and subsequent text composing can be conducted, be it for prints or screen rendering.

Have you included the question of character encoding in your project plans? I ask that because you didn’t mention it so far. But since you are about to introduce a software for recreating old sheet music prints, I can hardly view this disregarding the encoding matter.

The background of my remarks is my long-term project work on the Paulsmeier/Notationskunde edition (Schwabe Verlag Basel), for which I do the layout, composing and font development work. Karin Paulsmeier is the world’s leading scholar on early musical notation, her treatise on the subject comprises the notation from the 12th to the early 19th century, in three volumes. The typography is based on Andron (which is based on Renaissance type models) and hence can accomodate the special musical glyphs rather well. For the composing of this volumes more than a Hundred of special characters are required (and designed by myself) and here I face the same conundrum: only ca. 20% of them I can base on standard Unicode values, the rest of it run as a hack, because there is currently no other way. But, as I said above, if any beginning for an encoding of musical characters would be feasable at all, it would be the well-defined music lettercase of the Renaissance. Maybe there lies a great chance to boost musical typography into our century.

A few (not brillant) views of the Paulsmeier works, f.y.i.:
https://www.behance.net/gallery/17205603/Andron-in-print-part-1

Sam Gossner

Andreas Stötzner said:

There seem at least two sides of the challenge you venture for (which I find admirable). A.) the mere optical aspect of reproducing the appearance of the old printed sources; B.) the underlying system of the characters used for this kind of music composing, which has been based on a special lettercase’s content in the old days, but this system has no equivalent so far in our modern text reproduction architecture, which consists of character encoding, composing engine and suitable fonts. Which leads inevitably to the question: on which sort of character encoding your development is based. I presume it is a proprietary hack solution (similar like Sibelius or Finale deploy, to the day), please correct me if I’m wrong. That would enable to reproduce the optical surface of the sheet music in question, but not to convert the semantic content of those prints into actual transferable text information in the sense, in which we deal with all sorts of other texts, based on a standard character encoding.

It has been a failure to day to develop a viable composing system for sheet music based on systematic character encoding, mainly due to the fact that (western) musical notation is apparently too complex, similar as ancient Egyptian or Mathematical notation, which also embody a multi-dimensional text structure opposite to our common language-script-notations which are (more or less) mono-dimensional linear text-string systems.

Having said that, I’d like to point out if only one of the manyfold musical notation systems out there would be suitable for transmission into our modern technology, it would certainly be the lead-type composed music of (mainly) the Renaissance era. Because here we have a definite limited set of descret characters – and that is the main condition for defining a standard character encoding scheme on which fonts can be built and subsequent text composing can be conducted, be it for prints or screen rendering.

Have you included the question of character encoding in your project plans? I ask that because you didn’t mention it so far. But since you are about to introduce a software for recreating old sheet music prints, I can hardly view this disregarding the encoding matter.

The background of my remarks is my long-term project work on the Paulsmeier/Notationskunde edition (Schwabe Verlag Basel), for which I do the layout, composing and font development work. Karin Paulsmeier is the world’s leading scholar on early musical notation, her treatise on the subject comprises the notation from the 12th to the early 19th century, in three volumes. The typography is based on Andron (which is based on Renaissance type models) and hence can accomodate the special musical glyphs rather well. For the composing of this volumes more than a Hundred of special characters are required (and designed by myself) and here I face the same conundrum: only ca. 20% of them I can base on standard Unicode values, the rest of it run as a hack, because there is currently no other way. But, as I said above, if any beginning for an encoding of musical characters would be feasable at all, it would be the well-defined music lettercase of the Renaissance. Maybe there lies a great chance to boost musical typography into our century.

A few (not brillant) views of the Paulsmeier works, f.y.i.:
https://www.behance.net/gallery/17205603/Andron-in-print-part-1

Andreas, thank you for this incredibly interesting and insightful post. I will have to dig into this more later when I have time to really think more about what you have to say.

For now, let me iterate that we actually chose to avoid using any sort of existing typographic format for the musical notation.

I come from the background of sample library development, where we take analog, real instruments, and turn them into discrete, virtual instruments. As such, we based the digitization and organization of early notation characters on the way we would organize sample libraries, with each character 'sampled' multiple times, both as 'repetitions' and as 'variants', such as a character having a shorter stem when on the middle line. A JSON instruction file then explains to ENT what each 'sample' is and how it should be used/interpreted, and ENT then uses that to render the output, such as making sure characters are offset correctly and switching between variants depending on position. When a character is placed, ENT will select the nearest discrete staff line length (at specific ratios based on measurements I made of originals) which fits that character and combine the staff lines and character into a unit.

All this is done in the background without any need for the user's involvement, as ENT is entirely GUI-based.

Here is an example of one of the character sets in development; you can see this one has different shaped flags depending on line or space location, and the legs of the C clef are cut off when it is not in the center of the staff. You can also see there are variants for most note types (random variants are not currently implemented) which will serve to help break up the monotony of the page:


[[Edit: This process allows me to make a musical type font in about 1-2 days of light work, and allows some things which would take hours of work to implement in a normal font file to be done with the flip of a bit]]

The actual way we save entered data right now is in JSON format. ENT 'thinks' about each symbol only as a block of type; it is not aware, for example, that a semiminim is [usually] half the time value of a minim, it just knows it is a note block of value '3'. In the future, we would like to allow ENT to export in standard formats such as MIDI or musicXML, but making sure it is able to translate metrical relationships like tempus perfectus and diminution effectively is problematic.

If you have not already, you may wish to check out EMS Serenissima, which tackles the same issue but as a desktop font for use in Word and other conventional typesetters by using ligatures to essentially matrix inputs:
https://www.earlymusicsources.com/more/font-serenissima

To be clear, the Early Notation Typesetter is currently live and fully functional, and has been since the fall of last year. You can visit the website here and use the fully-functional demo if you wish:
https://typesetter.earlynotation.com/
https://editor.earlynotation.com/
(I chose not to post this information in the job listing so as not to accidentally risk contravening any rules against advertising products)

Text fonts, which is the intended topic of this thread, will be handed in the conventional way. My plan is to use Unicode when possible, plus MUFI for characters and ligatures which do not have conventional Unicode designations.

John Hudson

But, as I said above, if any beginning for an encoding of musical characters would be feasable at all, it would be the well-defined music lettercase of the Renaissance.

Do you envisage such an encoding cover the Renaissance music case as a glyph encoding, e.g. separate characters for each note on each position of the stave, with stave lines included in the character? And spacing elements of various widths, also with stave lines?

Just because a set of glyphs is well documented and able to be laid out without the complexities of later musical notation (or, for that matter, the combination neumes of earlier notation), does not mean that is obviously the way to encode those as characters. Would such an encoding be a ‘beginning’, or might it be a dead-end?

Andreas Stötzner

I fancy it would be a sensible beginning, but I’m not sure about it should be a dead-end.

I think the early music lead type character set would make a good candidate for standard encoding because it is a well-documented set of definite entities used for text composition of a considerable period; it may well count as a “script” or “notation system” in the way ISO and Unicode usually deal with these terms. The distinguished and remarkable achievement of lead-type music reproduction was, to break down a very diverse and manifold practice of hand-writing based musical notation to a limited set of a few dozens of lead pieces, accomodated in one single case, to facilitate music prints out of it for about two centuries. This notation system is, despite all similarities with e.g. Byzantine or Neume or later (modern) music notation, a script system of its own.

For me the overall question is: will musical notation ever move towards a standard encoding solution at all. Because its protagonists seem to live without it rather well, for quite a long period of time already. Sibelius and Finale have left the barn long since and more recent tool developments (like the ones Sam referenced above, e.g. ENT) again prove that a ‘dirty’ yet pragmatic way of making sheet music composing feasable, actually doesn’t neccessarily require a proper encoding (in the Unicode sense).

In the case of the old Renaissance lead type music lettercase (I can’t produce a representative image right now but I have it somewhere), there is good reason to regard this a definite character set as it is, that means a brevis at a-position is another character than the brevis on the c-position, and so forth. Because that composing system works that way. With the later and much more complex music engraving system, this is another matter and I presume that it will never become properly computable because no one will invest the effort neccessary to make this happen.

As long as you just want to prepare good-looking prints, all is fine with whatever hack one might utilize. Who cares? – But to make written/printed music compositions readable text in the computer-sense of the word, that is a challenge I don’t quite see anyone to cope with it to the day. The overall matter is too complex, it would require immense research and development endeavours.

Jeff Kellem

Andreas Stötzner said:

For me the overall question is: will musical notation ever move towards a standard encoding solution at all. Because its protagonists seem to live without it rather well, for quite a long period of time already. 

Well, there is the SMuFL (Standard Music Font Layout) private use area encoding specification that came out of Dorico development. Both Finale and Sibelius intend to support SMuFL, eventually. Lilypond can use fonts with those encodings. MuseScore uses SMuFL fonts and intends to support 3rd party fonts in a future version. So, there's hope (for being able to potentially use a single music notation font across scoring software).

John Hudson

My experience with fonts for mathematics—which have some similar levels of complexity in arrangement—makes me think that fonts are too low level to be trying to solve music notation. At most, fonts will be able to provide glyphs and potentially some data to be used by higher-level layout engines. So what is needed is the full stack of a common mark-up language—something like MusicXML?—, Unicode encoding of more-or-less abstract characters, a well-specified music layout engine, and then the fonts. And if you have all that, it seems to me that early Renaissance notation could be fairly easily accommodated in that stack, rather than treated in its own way as a glyph encoding for text-like reproduction.

Which is not to say that I don’t think a proposal such as Andreas suggests wouldn't have any merit. I would read it with interest, and may be convinced.

John Savard

John Hudson said:

My experience with fonts for mathematics—which have some similar levels of complexity in arrangement—makes me think that fonts are too low level to be trying to solve music notation.

Mathematics and music indeed use two forms of notation that are more complicated than text. They both use a set of glyphs that can be handled in the usual manner (and, in fact, people made "music typewriters"), but the positioning of those glyphs is more varied.

However, there's an important difference between the two. One could imagine a system of math notation that looks like this:

.int(a,b,/(^(x,2)+x,.rt(x+1)),x)

to specify a definite integral; a, b, the expression, and the final dx all "hang" from the integral symbol in known places.

In music, there are different symbols for the whole note/semibreve, the quarter note, and so on, but the vertical position of the note on the staff indicates the actual note, so one has a lot of symbols that have to be accompanied by positioning information.

So if one wanted to devise a system of typing music from a conventional keyboard, as opposed to painting it with a mouse, one would have to type position codes as well as symbol codes (plus, of course, grouping notes to be played together).

So while I definitely agree with you that fonts aren't enough to 'solve' even mathematical notation, most ordinary mathematical notation needs just a little bit extra compared to what music needs.

Sam Gossner

ENT is 'pitch agnostic'; it only knows which staff line/space it is on a integer +/- basis with the middle of the staff being '0'. Part of this is due to the much larger range of clefs in common use in the period.

Some of the earliest attempts at digital music typesetting did in fact use a direct system of typing music via conventional (text) keyboard (e.g. SCORE) and this system has proven itself very well in many publishing houses over the decades as efficient and surprisingly capable in experienced hands. While most composition and sequencing software today relies on MIDI data as a backbone, allowing for very fast entry with the use of a MIDI controller keyboard, some could make an argument for an effective (text) keyboard-only method.

However, in my opinion the ultimate issue with that comes down to the learning curve and accessibility. Having users type in note names and pitches requires a mental translation layer of visual -> semantic; no problem for an expert, but a student or non-musician will be very slow or completely shut out from the process. That's why ENT is designed to work with visual pattern-matching, and it can accept both fully-keyboard and fully-mouse/touchscreen input, or any mix of the two methods. In this way, the entry method can be customized or altered during entry to best suit the circumstances being typeset. Ultimately while we want to reflect the typesetting origins, we also want something which is hopefully faster than real-life manual typesetting.

Regarding music typesetting and the possibility of a singular standard, one of the main issues we see in that area is that different people want to achieve different things with notation. It's like having different factions of mathematicians which use different symbols and methods to accomplish similar (yet distinct) tasks. While SMuFL does a venerable job attempting to cover all the necessary characters required by a notation program to spit out something nice in the end, to create a fully-compatible font with all extended capabilities (thousands of glyphs) would be a massive undertaking, not to mention there are users who would still be dissatisfied with the lack of certain esoteric glyphs used in their area of music notation.

The music industry as a whole is a deeply fragmented place. There are almost no universally-agreed standards aside from MIDI and PCM WAV audio, without which collaboration would be nearly impossible. Software cannot even agree on a pitch standard for describing which octave a note is in (C4 = MIDI note 60, C3 = MIDI note 60, and C5 = MIDI note 60), much less how to encode notation in a consistent way accessible to all programs.

My primary goal with this project is not to provide open exchange with other software, but to provide accurate visual results in a fast and repeatable manner. While exchange formats like MIDI or MusicXML are in the cards, we still need to finish implementing essential notation features and text before we can go that far.

Andreas Stötzner

Jeff Kellem said:

Well, there is the SMuFL (Standard Music Font Layout) private use area encoding specification that came out of Dorico development. …

SMuFL is a great achievement, no doubt, But the fact that this encoding scheme does not (yet) work with some of the major music engraving applications, shows the grave difficulty of progressing with a wider implementation of such an ‘insular’ encoding. I am familiar with all the pros and cons of a PUA-based strategy, from my involvement with MUFI, mainly.

However, in terms of “standard character sets“ – or what one may consider as such –, the field of musical notation is very much complex, indeed. And the sub-field of lead-type character sets is a special case within that field – which is distractingly diverse in itself, which doesn’t make things easier.

I managed to unearth a few references from my archived materials, I just show quick preview shots in order to outline what falls into that category. I’m not aware wether some of it is known or not, the selection is rather accidental and by no means representative. But it illustrates some noteworthy aspects, I gather.

The two first images are taken from:

Joh. Heinr. Gottfried Ernesti:
Die Wol-eingerichtete Buchdruckereÿ
Nürnberg, Buchdruckerei Endtner, 1733
(from a reprint of 1940)

There are similarities and differences between the two sets. They demonstrate, however, that letterpress-composing of sheet music was quite common even past 1700, in fact it was still used until the 20th century (e.g. for church songbooks).

This is a much more extended lettercase scheme from around 1890. It obviously follows a different structural system compared to the previous examples; one might actually see it as a collection of glyph components. The illustration is from:

Alexander Waldow:
Lehrbuch für Schriftsetzer
Leipzig 1888

and credited for it is the Julius Klinkhardt book and music printing officine in Leipsic.

Back in time and entering an even more specific chapter of musical notation and its rendering by means of discrete types, is this example of a tablature notation (which is ascribed to Alessandro Piccinino):

Tablatur notation was very popular for a long period of time, mainly for chord instruments, but also for organ. It is a totally seperate sort of music notation since it depicts not tone/length on a staff base but finger indications related to the strings of a lute, chitaronne, guitar, violon or similar instruments. In the sample above the tablature is accompanied by ‘ordinary’ notes, but that feature is not mandatory for this kind of notation. If I am not mistaken, the single unit of this setting is not the complete piece of the staff with 6 lines, but rather the single figure stroked, and these bits are composed together. So, what is the glyph, what is the character here?

Craig Eliason

Lines of staves are “justified” for the sake of aesthetics and function, aren’t they? Another reason to solve the layout challenges at a higher level than the font. 

konrad ritter

I love the Fournier italics you made after Bernoulli's Opera. (Is it Jakob or Johann? I rely on those editions a lot, for my work--totally unrelated to yours.) Is your italic for sale? I'm going through a major Fournier phase these days, and I'd love to use it for my papers. Nothing commercial, to be sure. 

Sam Gossner

konrad ritter said:

I love the Fournier italics you made after Bernoulli's Opera. (Is it Jakob or Johann? I rely on those editions a lot, for my work--totally unrelated to yours.) Is your italic for sale? I'm going through a major Fournier phase these days, and I'd love to use it for my papers. Nothing commercial, to be sure. 

I'm afraid not, it's actually missing a ton of characters still; sadly only a few pages of italics to work from! That's why I'm posting here, to find someone willing to help fill in gaps like that.

I made it from photographic scans I took of an original book (a relative of mine was an astronomer who studied the history of astronomical phenomena, in particular eclipses, and left behind a number of books). If you would like access to the raw photographs to make your own perhaps, or a copy of what I have so far (nothing fancy, just auto-traces at this point for experimentation), send me an e-mail or direct message.

Andreas Stötzner said:

I managed to unearth a few references from my archived materials, I just show quick preview shots in order to outline what falls into that category. I’m not aware wether some of it is known or not, the selection is rather accidental and by no means representative. But it illustrates some noteworthy aspects, I gather.

The two first images are taken from:

Joh. Heinr. Gottfried Ernesti:
Die Wol-eingerichtete Buchdruckereÿ
Nürnberg, Buchdruckerei Endtner, 1733
(from a reprint of 1940)

There are similarities and differences between the two sets. They demonstrate, however, that letterpress-composing of sheet music was quite common even past 1700, in fact it was still used until the 20th century (e.g. for church songbooks).

This is a much more extended lettercase scheme from around 1890. It obviously follows a different structural system compared to the previous examples; one might actually see it as a collection of glyph components. The illustration is from:

Alexander Waldow:
Lehrbuch für Schriftsetzer
Leipzig 1888

and credited for it is the Julius Klinkhardt book and music printing officine in Leipsic.

Andreas, these images are remarkable! Thank you for locating and sharing them, they are incredibly informative.

I put together a 19th century variant of a typeset music primer from about 1880 (an Oliver Ditson publication from right here in Boston, if I recall) for ENT which is quite fun to play around with, especially when in 'uninterrupted staff lines' mode:


Like in your example, the whole notes are extra-wide. It seems by this period though, that it was more common to construct the staff out of more parts rather than the fewer, larger chunks of the Renaissance, as you say, glyph components; even the flags [edit: and stems!] are unique components. Examining the 'break' patterns show the difference:

A fair guess might be that the increased complexity of having to notate complicated ties, beams, and an ever-expanding range of articulation symbols perhaps demanded a system with greater flexibility and less parts.

Craig Eliason said:

Lines of staves are “justified” for the sake of aesthetics and function, aren’t they? Another reason to solve the layout challenges at a higher level than the font. 

Yes, though at the end of the piece they would usually fill the line so as to justify the last line with a similar spacing to the others, and possibly fill out the rest of the piece with blank lines too, as shown below.


Pub. Paul Ledertz, 1623; PMLP1014561

The exact specifics of how and why lines were spaced varies massively between publishers, including where extra spacers are usually allocated and what width of spacer is used in what contexts. What is certain is that it differs significantly from modern aesthetics of music spacing, even though the underlying rules (fit text underlay/lyrics, avoid clutter around important symbols, keep lines justified evenly) are almost the same.

John Hudson

So if one wanted to devise a system of typing music from a conventional keyboard, as opposed to painting it with a mouse, one would have to type position codes as well as symbol codes (plus, of course, grouping notes to be played together).

It is important to remember that input ≠ encoding. There are multiple text representations of music notation already in use in tools—Lilypond’s syntax, MusicXML, SMuFL—but no recognised standard that works everywhere as a common storage and interchange encoding. [That is also true of mathematics typesetting systems, but there is at least a W3C MathML working group effort.]

If there were a common standard for e.g. MusicML, the input method sitting on top of that could vary widely.

[A few years ago, I typeset a lot of Gregorian notation using a fairly simple text representation that worked with a particular font using OpenType GSUB to form neumes based on recognised relative pitch patterns. That was entirely font-specific, and that seems to me the problematic aspect of any direct character-to-glyph model, because it seems to me desirable to encode a Gregorian melody and then be able to present it in Gregorian or modern notation without needed to re-encode it.]

Scott-Martin Kosofsky

Music type was always a commercial compromise, the fastest method for reproduction of simple notation in the early letterpress era. It was adequate for much 15th- and 16th-century part music and simple tablature, but by the 1580s, as instrumental music became more complex, it was no longer practicable and was surpassed by intaglio engraving, which could be both clear and beautiful. (This marked the beginning of modern musical notation.) Music type did improve, as seen in the German examples posted above, reaching an apogee with the music types made in the mid-eighteenth century by J. M. Fleischmann for the publisher J.G.I. Breitkopf. But still, when compared to engraving, it was (to adapt from Dr. Johnson) “like a dog's walking on his hind legs. It is not done well; but you are surprised to find it done at all.” Later, in the 1830s, lithography became a favored medium, but was no aesthetic match for engraving.

Sam, the overriding question, to me at least, is “why bother?” And, especially, why bother with a system, clever as it may be, that cannot be transferred easily to any other music composition system? This is the era of portability. The need for fonts with Renaissance note heads and symbols cannot be denied, but why reproduce the less sophisticated ones with annoying joints in the staff segments? To reiterate my comment above, the noise-to-signal ratio is just too high.

Sam Gossner

Scott-Martin Kosofsky said:

Sam, the overriding question, to me at least, is “why bother?” And, especially, why bother with a system, clever as it may be, that cannot be transferred easily to any other music composition system? This is the era of portability. The need for fonts with Renaissance note heads and symbols cannot be denied, but why reproduce the less sophisticated ones with annoying joints in the staff segments? To reiterate my comment above, the noise-to-signal ratio is just too high.

Regarding why I am doing this project:
The fact that this form of notation is not [easily] accessible in conventional music notation software is a reason enough for me. Try spending an afternoon satisfactorily re-typesetting 16th century music in Finale, Sibelius, or any other notation program.

The problem is, all of the modern notation software solutions are designed around engraved, measure-based notation, and a reflection of 19th and 20th century ideals and practices. The minute you place a note which goes over a barline, it will complain, spacing will break, etc. Even if you spend hours tweaking to remove barlines, ignore errors, or add editorial compromises such as bars with extra beats, the intention of the music has been irrevocably altered. I won't deny there are some solutions which can get closer with more respectable spacing, but in all of those cases, it is using modern symbols, straight, digitally-perfect stems, and smooth, modern noteheads... a compromise almost worse than either-or in my opinion!

To use such software to do a clean, fully-modernized retypesetting (barlines with tied notes, modern clefs, no custos, nothing fancy) is fast and relatively easy, and the result is reasonably useable, just not historical: fine for students, hobbyists, and maybe even some academics, but certainly not everyone. To use such software to make an even halfway decent approximation of early notation (no barlines, no ties, custodes on every line, historical clefs, "historical" spacing, musica ficta, etc.) is an exercise in frustration and hours of manual tweaking, 'hacking', and generally trying to fit a square peg in a round hole.

The closer you want to get to true early notation in a notation/engraving application, the more painful the experience becomes. Want to represent longas and maximas? How about ligatures of any length and consist? Half-coloration? Sign of congruence? Dots of division? Forget. It. You'll have wasted 20 hours on hacks by the time you get anything remotely attractive done; the software just is not designed for the task.

Dozens if not hundreds of people have been struggling with and trying to solve this problem for over 40 years, and to me the attempt to 'hack' the solution into an existing typesetter or notation program is a part of why it hasn't worked yet; it's like trying to make a high-end CD player read shellac 78 rpm records, when you should just be using a dedicated turntable... and yet you are asking me why bother using a turntable to play all these 78's when I can find some modern recording of someone else playing the same song.

To me, early notation is designed for this music, and this music is designed for early notation. You can separate them, but the minute that is done, you have lost part of the message, part of the artist's original intention. Just like the artists of early Jazz who put their sound on wax and shellac, we no longer have the people responsible for the music of the Renaissance with us; we must rely on every context clue they provide us.

And frankly, I think the claim that this notation is "less sophisticated" is alarmingly reductionist. It was the system of its time, and as such, it provided every affordance it could to its users. Indeed the number of glyphs available was less, but it sure as heck didn't stop them from writing some of the most complicated and diverse music of the past millennia. Frankly the talent and capabilities of the musicians and composers of that period were just as incredible as any other period in modern history.

Even though my experience with reading modern to early notation is probably a 100:1 ratio, I still find it easier to read, say, a Gastoldi duet in its original form off a facsimile than a modernized edition (or worse, semi-modernized). The shape of the notes with their tapered stems, the way they are spaced and organized, the particular weights and angles of the glyphs, and the condensed, compact format; it is conductive to reading the particular musical shapes and patterns of this period, in the same way our modern notation system is designed for notating the particular shapes and patterns of 19th and early 20th century music. Maybe it does not live up to our modern concepts of "beauty" or "aesthetics"? If so, why do we bother keeping old buildings or paintings around?

That's not to mention the fact there are thousands of scans and original manuscripts in such poor condition or quality that they are virtually unreadable, unplayable. No concert musician can perform the pieces in their current state, so it is valuable to be able to re-typeset such pieces in a "clean" but respectful format.

After all, are there not countless efforts to make clean, modern facsimiles of ancient scripts possible, be it hieroglyphics or Linear B? In what way do you find this endeavor different or of less value? This music reveals countless glimpses into the lives of people who lived, worked, and died hundreds of years ago and deserves to be explored and enjoyed in the most authentic way possible.

Yes, one could "hack" modern notation software for many hours to get an output which is marginally authentic. Yes, one could give up and make a fully-modern edition and say "forget all those old funny characters," but in every change, part of the message gets obfuscated.

And, lastly... why not? "Accessibility drives preservation." Not my words, but rather those of Brewster Kahle, founder of the Internet Archive. Worst case, one single work or composer is raised from obscurity, one person decides to research this subject more, one person finds music they enjoy. How can that NOT be worthwhile?

I hope this addresses your question "why". It is sadly not a short or simple answer, but my attempts to explain this to you before apparently were insufficient. If you wish to discuss the specific matter of why in greater detail, may I suggest we discuss it over direct message so not to stray so far from the topic at hand.

Regarding your other concerns:
1. We will add some form of export option, perhaps MIDI or MusicXML, just doing that right now is massively putting the cart before the horse... metaphorically the horse isn't even old enough to pull the cart in the first place. Don't get me wrong, this will be a challenging process, but the rules for note ratios and guidelines for converting this notation into modern forms are well documented and well understood.

2. As I have mentioned several times before, the form and appearance of staff segments are completely customizable by the user and can be merged together to form smooth staff lines if desired (such as for triple-process notation a la Petrucci's Odhecaton). The staff lines are a separate element from the glyphs and are adjusted in their own way globally, as shown below:

John Savard

As for fonts with the extra characters used in older printing, have you looked into Cardo and Junicode, among those mentioned on this site:

https://folk.uib.no/hnooh/mufi/fonts/

Also, someone else appears to be doing work similar to yours:

https://casfaculty.case.edu/ross-duffin/homepage/fonts-for-early-music/

I certainly understand that musical notation has changed significantly over the years, and thus software designed to typeset modern musical notation would not necessarily be suitable for, say, the notation used for Gregorian chant (which is standardized, and still in use)

let alone the version of notation used for ordinary music in the 16th century or even in a 17th century document like Simpson's Division-Viol.

Although, as this sample from that document shows:
by then, music was already being divided regularly into measures, and I also recognize the clefs and the key signatures; only the time signature seems to be missing.

Sam Gossner

John Savard said:

As for fonts with the extra characters used in older printing, have you looked into Cardo and Junicode, among those mentioned on this site:

https://folk.uib.no/hnooh/mufi/fonts/

Also, someone else appears to be doing work similar to yours:

https://casfaculty.case.edu/ross-duffin/homepage/fonts-for-early-music/

I certainly understand that musical notation has changed significantly over the years, and thus software designed to typeset modern musical notation would not necessarily be suitable for, say, the notation used for Gregorian chant (which is standardized, and still in use)

let alone the version of notation used for ordinary music in the 16th century or even in a 17th century document like Simpson's Division-Viol.

Although, as this sample from that document shows:
by then, music was already being divided regularly into measures, and I also recognize the clefs and the key signatures; only the time signature seems to be missing.

I'll look into those, thanks for the suggestion. I'm quite familiar with MUFI, it's been a great help with some of the more unusual ligatures and characters in the fonts I've been assembling for the project. We don't need anything close to thousands of glyphs for this, but there are a few used in text underlay (lyrics) which are to the best of my knowledge not represented in MUFI or Unicode, so those would need to be added one way or another if the author permitted. Either way, something to explore.

Yes, Ross Duffin has actually been working on retypesetting early notation since the 80's; he's come up several times now in this discussion and I am very familiar with the existing work in this area and his work in general. Early music is a relatively small community, and this project has been public for about half a year now, so these are things we've discussed quite a bit so far.

Interestingly there is a Gregorian chant notation program out there which is apparently quite decent and has been used in almost exactly the same use case as ENT is designed for (retypesetting damaged or otherwise inaccessible works). ENT is focused rather on what is termed 'typeset white mensural notation', a form of notation and printing method used between about 1500 and 1650, while Gregorian chant is from centuries prior.

Barlines at regular, metrical intervals started to appear near the end of the 16th century and grew in popularity through the first half of the 17th as the style transitions from Late Renaissance to Early Baroque. "Score" format works, where multiple lines are interleaved row by row, almost necessitate their use, and indeed some of the later fonts in the ENT project are from such sources. Not sure why no time signature is given in that particular work, but time signatures have been around since at least the start of mensural notation, so perhaps the composer just assumed the performer knew from looking.

Right now we're focused more around 1530-1600, where barlines were not used as they are today. We also do not plan on adding engraved or hand-drawn characters such as in your example, at least not for a few years.

The whole point of this is to make something that is intentionally 'dumb', basically a glorified, specialized word processor, so that it allows a wide range of inputs and options without frustrating the end user over preconceived notions of what it is being used for. At the same time, there are features which help to ease the more tedious aspects of the task, and a significant amount of control over the appearance of the final output, so the users themselves can decide what specific publisher style of notation and staff appearance they wish to use, all with a few clicks.

Andreas Stötzner

“…there are a few [glyphs] used in text underlay (lyrics) which are to the best of my knowledge not represented in MUFI or Unicode”

can you specify which characters or glyphs you don’t find in neither MUFI nor Unicode?

Dan Reynolds

Scott-Martin Kosofsky said:

Music type did improve, as seen in the German examples posted above, reaching an apogee with the music types made in the mid-eighteenth century by J. M. Fleischmann for the publisher J.G.I. Breitkopf.

My comment is a little off-topic, but Fleischmann did not make musical notation type for Breitkopf. Rather, Breitkopf’s printing house included one of the larger German typefoundries of its time. Breitkopf had his own punchcutter cut the punches according to his instructions for a musical notation font his foundry produced, with a method of setting he devised. Fleischmann then copied this (I think for Enschedé), much to Breitkopf’s chagrin. In the 1770s, Breitkopf published an essay that more-or-less condemned Fleischmann and Enschedé. The music-copying episode was just one of the points of criticism he levied.