These music typewriters are certainly interesting curiosities but never really caught on like the advertisements and curious articles would make you think. Music notation is has far more graphical needs than words as the horizontal and vertical spacing of objects have meaning beyond the objects themselves. So spacing on a typewriter like this can be extremely difficult to do well. This is why metal plate engraving was the dominant form of music engraving for a final published product right up until the 1980s when computers took over. It was far more reliable and you could get any level of complexity of notation. It did the same thing as the typewriter, but the human was the typewriter rather than a mechanical contraption.

While the metal-side of the art is lost, the musical and graphical skill required to make the notation still lives on in "engravers" today who do the work on computers. Practically all music notation today is done on computers with the various softwares and platforms out there. Rendering good music notation is a very difficult problem to solve but most of the good engines have finally gotten to the point where any reasonable notation is possible, and even some unreasonable ones. There is little love lost for the old methods of engraving in my generation, especially given how expensive and time-consuming it was compared to today.

Full disclosure, I'm on the Sibelius dev team and have used it for close to 20 years, so I definitely have a horse in the race of which platform is best. But it's clear that with more competition in this small market, each platform is pushing the others to get better and more flexible.

I've used several different notation programs over the years and most of them make me want to scream and cuss over how un-intuitive they are compared to hand-writing a score. There is so much complexity and specificity required of the tools.

No doubt you've probably heard of Tantacrul and his Youtube channel. Watching him pick apart the design issues of every major notation software really highlights just how hard it can be to get this right. His channel occupies a nice niche at the intersection of music composition, design, and programming. Also his videos are entertaining as hell.

I particularly enjoy his video about the making of a new notation typeface. He does a good job at making the topic of music notation come to life, and also throws in a short history of the program 'Score' for some context: https://www.youtube.com/watch?v=XGo4PJd1lng

There may be little love lost for the old methods, but there is much love lost for many of the results.

Here's a talk I gave about building its music graphics engine in client-side JavaScript, with an emphasis on performance hacks: https://www.youtube.com/watch?v=XH5EtQge_Bg

Nowadays, I'm so used to my new workflow of writing a script that checks my music, creates a LilyPond file (as well as a .musicxml to work in MuseScore4), and generating a PDF using lilypond. There is just something so uniquely beautiful in old school sheet music -- we composers consider sheet music notation as its own art form. There are tons of strict rules that you need to follow (to make it readable to the musician) but there is also tons of creativity you can pour into your notation. It's like programming a real human-being's finger motions! Some contemporary composers (such as imho Unsuk Chin) contributed so much to the art of music notation, as well their contributions to the art of music. When I'm done with a composition, generate the PDF, print it, and hold it in my hands for the first time (many fixes to be still made on the first draft!) it's the happiest moment. I can only imagine how hard and time-consuming it was to create draft after draft with this typewriter back in the day. I can only imagine someone working day and night transcribing every work of Bach on this typewriter. Must have taken many man-months maybe man-years! Us new generation of composers are so spoilt with the availability of insurmountable amount of software tooling around music.

My intuition is that drafts would be made by hand and the Musicwriter work would be done as transcription in preparation for printing…usually by someone else within an infrastructure of editors, publishers and printers.

Not that I wouldn’t bet some composers had their own Musicwriter, but the Musicwriter hails from the days of secretaries and typing pools.

Today people typeset their compositions because they can. The goalposts have moved in recent decades. For better or worse.

Can you expand on this a little bit more ? I hate my workflow for scoring, and I'd like to have something a little more efficient for generating multiple parts/etc.

- In MuseScore create way more parts than you need. This serves as a temporary work space.

- Save tons of files. E.g. if I'm working on "Composition_00341" I save bunch of files like "Compo_00341_{DATE}_AUDIO", "Compo_00341_{DATE}_ALL_PARTS", "Compo_00341_{DATE}_VIOLIN_I", "Compo_00341_{DATE}_SLIDES". They all have different purposes, such as one having optimized audio playback, the other for all parts, others for single parts, and one for 1920x1080 screens (to make a score video), etc.

- Use your notation software for your benefit. If you need more space, use large paper sizes (such as A0 etc), add more parts, add colors, add notes, leave more space between parts for more notes, add more measure numbers (so you can reference them in your notes) etc... I also use quirky, personal notation for my own purposes that I remove out in the final draft. E.g. I don't like seeing 8va, 8vb lines, or alto clef, so I only use F and G clefs with 8 or 15 on-top/bottom to simulate 8va/8vb/alto clef. Ultimately, these need to be removed since musicians prefer 8va/8vb/alto clef (e.g. for viola). Also some musicians (such as flutists) prefer seeing tons of ledger lines instead of 8va, so you need to think about these things, but while I'm composing I want to see an octave G-clef (G clef with 8 on top). It's easy to fix these things in MuseScore with a few clicks.

My new workflow is like this, most of this is still work in progress but I really really like it as it makes me very productive not just as a composer but also as a notation engraver:

- Custom python script. Use musicxml [1], mido [2] and lilypond [3]. They all serve different purposes.

- First, I start my compositions with sketching some ideas in MuseScore, or real-life in piano, this is just the old-school centuries old workflow. As most musicians would agree, usually there is no replacement to playing an idea on the piano, or sitting down and listening to it in your head. Philip Glass says that making music is the same activity as listening to music (except in your head!).

- Second, write parts of the composition into this script. I like this script checking certain things I don't want to make mistakes in and are too time consuming to check one note at a time (which is likely the best way to do this). For example, counterpoint. I also (occasionally) use 12-tone rows (or other forms of Serialism) and this can be checked too. Ultimately, these matter as much as you want them to matter. I also like using some "algorithmic composition" ideas, to quickly mock up some form I think about and generate a composition with that to test it out. You can use it as a battleground of sorts.

- Third, this script generates 3 main things:

- - A MIDI file for audio playback. Now, I personally think MIDI is absolutely horrendous and imho anything other than keyboard instruments (namely piano, organ, celesta, and harpsichord) sounds unusable (unusable = more confusing than not listening to it in the first place). This is fine for my purposes, as you can work on orchestration later in MuseScore/lilypond. This is just the pre-alpha draft.

- - A lilypond file. This generates a PDF. This can be the final PDF you give to the musician(s), or you can generate it through MuseScore.

- - A musicxml file. This can be used to import your music into MuseScore. You want to do this if you further want to work on your notation in MuseScore, or use MuseScore playback. Another option is to customize the lilypond file. Now, I know that some composers prefer using lilypond, I personally like MuseScore better.

[1] https://pypi.org/project/musicxml/

[2] https://mido.readthedocs.io/en/latest/

[3] http://lilypond.org/

I write a lot of multi-part scores (i.e for big band and small jazz combos) and the idea of linting to check for things like note doubling or parallel fifths is a dream of mine.

When you say write parts of this composition into the script - what form does that take ? Is it midi ? Is it score data generate by a musescore plugin ? This workflow sounds amazing and I'm really curious how you get from point A to the intermediate point.

It's a custom programming language. I have many designs for it, but unfortunately almost nothing is implemented yet. Only exactly what I need is implemented. Currently, it looks very similar to lilypond language with some lisp-y compute help because lisp is easy to implement. E.g. I can do:

  (set melody1 "c' d' e'")

  (set melody2 "g' g' b'")

  (set diff (- melody2 melody1))

  (print diff)

  (print (to-intervals diff))

  [7 5 7]

  [P5 P4 P5]

I like using serialistic techniques so I use this to check my rows, also note when I'm violating rows. Another thing I do is to highlight parallel perfect fifths in red so that I can change them if I desire. I also detest vertical minor seconds, and highlight them in red too just in case I accidentally add a minor second between parts. I also think major/minor thirds below E below mid C are too muddy on piano, so highlight them as well.

[0] https://www.humdrum.org/

I need this because, sometimes I want auto-generated annotations to my score. Say I write a phrase in MuseScore, I'm interested in the script adding a text with chord names, or pitch Set Class etc. MuseScore can do this with plugins, but I personally find MuseScore plugin system very buggy and hard to program. Python is much much much better and musicxml is pretty easy to deal with.