Why I'm looking forward to Musescore 4

• May 11, 2021 - 20:42

I've uploaded a Musescore project to YouTube, an original concert piece for contralto and orchestra called Vienne la neige. Because you can't get a computer to sing like a contralto, the vocal part is transcribed for cello.

It's hard to read the 10-point text in the score and the text is in French, so there are two versions: one with a moving cursor score, and one with text slides and a running English translation.

Text version:


Announcement (shoutout?) concluded, I'd like to say I'm very pleased by the direction Musescore 4 will be taking. Realising Vienne with 3.x was the most gruelling 18 months I've ever put into a piece. I set myself the task of realising the score entirely within Musescore--using the internal FluidSynth, no pay-or-you-don't-play commercial sample libraries, expression controlled directly from the score, phrasing/bowing and articulations from the piano roll editor, etc. Pure Musescore, in other words, except that the reverb is supplied by an external zita-rev1 reverberator, the same one Musescore 3 uses but operating in ambisonic mode, which Musescore doesn't provide.

It was an exercise in frustration from start to finish. If I were to grade Musescore 3, I'd give it a nice solid B+ for engraving, and a resounding F when it comes to realising orchestral scores. Fully half the time I worked on the piece was spent concocting hacks, kludges, and workarounds to playback/midi deficiencies and flaws. The absolute very worst of these was the incompatibility of "expressive" soundfonts (SND) with user-set note velocity offsets. Hundreds of Staff Text channel changes and uncountable numbers of hidden dynamic markings later, I was nearly weeping with frustration.

The promise of a Musescore that takes the needs of midi composers seriously, as 4 promises to do, fills my heart with joy. I want to compose, engrave, and realise scores, not fight with software for months at a stretch. Kudos and a huge thanks to Martin Keary for taking over Musescore. (His Shostakovich video alone convinces me he's the person for the job.) I am unreservedly optimistic.


Very nice work. But as a Msc "purist," will you continue to confine yourself to the HQ sf3 font? Dynamic expression writ large requires multisampling

In reply to by ramblinj

I normally don't use the Musescore sf3 soundbank. Even in this piece I didn't use it exclusively. The oboe, English horn, celeste, xylophone, and all of the string departments come from the MS sf3 sounbank; the remainder are from my personal library of free soundfonts, loaded into Musescore so only the built-in FluidSynth sampler is used. I usually send midi from Musescore to an external sampler, but I wanted to test Musescore's effectiveness as one-stop "composition software," i.e. one where the composing, scoring, previewing, and realisation were done entirely from Musescore. As my comments make clear, Musescore 3 doesn't cut the mustard, but 4.0 really looks like it will. I'm very excited.

I’m as delighted as anyone by the direction MS4 is taking, but there’s a formidable amount of new development involved and I’m trying to keep my expectations modest about the initial release. Why are you so set against using other tools? Cakewalk is a very capable free DAW, for example. I hope you won’t take this the wrong way, but the ordeal you describe sounds a bit self-inflicted.

In reply to by scott.1

I'm not a Musescore purist, as someone else suggested, and I'm not against using external tools in conjunction with Musescore. The book I wrote on Linux midi orchestration ( https://www.schaffter.ca/pdf/linux-midi-orchestration.pdf ) after MS 2 came out makes that clear. It was for this one piece only that I decided to see how effectively Musescore, by itself, could accomplish all the steps from composing to notating to realising the score. Think of it as "rigorous testing of a program." So yes, to some extent, the ordeal was self-inflicted.

A DAW like Cakewake is not a suitable program for composing and realising orchestral compositions, which require "conducting from the score," so to speak, not tinkering with sequencer tracks and piano roll editors. The creative process is entirely different from assembling tracks for a hip-hop song. The midi instructions for driving the sampler (I generally use the venerable sf2/sfz/gig-capable LinuxSampler) have to be embedded in the finished score, and Musescore is the only GPL'd notation program that comes close to providing that in any meaningful way. It just doesn't do it very well--yet. "Playback" has for many years been Musescore's poor relation. I'm assuming that will change with 4.0.

In reply to by Peter Schaffter

The trouble with “conducting from the score” - in my view, which I’m willing to reconsider - is that in many or even most kinds of music only a small fraction of the performance is actually encoded in the score: the rest is specified by tradition and context. A crescendo drawn across so many beats shouldn’t necessarily be rendered as a linear progression; staccato or accent marks may mean any number of different things, even from one measure to the next. When I think about the user interface needed to shape these things at the necessary level of detail, what I come up with is precisely “tinkering with sequencer tracks and piano roll editors.” It’s not fun, but is there an alternative?

I don’t do hip-hop, by the way.

In reply to by scott.1

Thanks for carrying on this thread. It's a useful discussion. I appreciate it. I don't do hip-hop either. :) I used it as an example of a genre that is ideally suited to the sequencer.

Printed scores encode only a portion of what is needed to perform them, true, but midi allows us to encode the rest. It's a great boon to composers like me who are perfectly aware that most of their works, particularly orchestral pieces, are unlikely ever to be performed. Securing orchestral premieres has ever been a chassse gardée.

It's a matter of personal preference whether the midi encoding is done "at the score" or "in a sequencer." The labour involved is the same. In both cases, you click an object (a note in a score, a tetris block in a sequencer--sorry, I just can't resist the quip) and add performance encoding. Perhaps it is merely a deficiency of mine that I need to see the score in order to make meaningful "performance" calls. I have yet to develop the mental faculty of "reading" sequencer tracks the way I read a conventional score, though not for lack of trying. I can't "see" bad voice leading, or inappropriate doublings, or form a mental image of the harmony, or trace the interaction of polyphonic lines.

While agreeing that note-by-note performance encoding is always painstaking and tedious, the alternative you ask about is, of course, to do it from the score. With a few truly lamentable exceptions, Musescore's Inspector paradigm makes that possible: click on a object (note, articulation mark, dynamic, etc.) and make performance adjustments in a conveniently located panel that doesn't occlude the score.

Perhaps one of the developers will read this, so I'll list the lamentables, which really need to be addressed in MS 4.

  1. The inability to change a note's velocity offset (i.e. make it louder or softer) when using SND soundfonts. This presents a nearly insurmountable obstacle to "from the score" performance encoding, as I discovered.

  2. The inability to change the "meaning" of articulation marks to fit the context (e.g. make staccati shorter or longer).

  3. The inability to control note on/off-times from the Inspector, which is essential to phrasing (winds and brass have to tongue or breathe, strings have to change bow, everybody has to be able to play legato or portamento, etc.). In a midi realisation, these can be very satisfactorily simulated by adjusting note on/off times.

In reply to by Peter Schaffter

It seems to me that the way MuseScore works is that the Inspector Is more for the engraving side of things, and the Piano Roll Editor is more for the way playback works.
A lot of playback gets muddled with too much reverb.

MuseScore seems to be more optimized for keyboard shortcut users. Poor mousers like myself are left out. Or told to just learn the shortcuts.

Some things I would like to see:

Select a range of notes and add an articulation. Later, be able to select the same range of notes, select the same articulation to delete them.

Have staff text like rit. affect playback a preset amount. Adjustable in the inspector.

Be able to put a note on any beat without having to fill in rests first. For mouse users mostly.

Add measures at the end of a piece by selecting the last measure and hitting "R".

Select a measure and be able to have that staff only start playback from that measure. Select a range of staves and have them play.

I can't think of anything I like about the drum palette.

In reply to by Peter Schaffter

Peter, I think we may not disagree about anything very significant after all.

I find the DAW very useful for certain tasks. I think the piano roll view is good for the “horizontal” dimension - seeing the durations of notes and the gaps between - but as you say, hopeless for anything to do with pitch. I also need to see a series of MIDI velocities (or whatever is controlling the loudness of a phrase over time) as a graphical shape that I can redraw with a sweep of the mouse. So to me, the DAW is where I can quickly select an arbitrary number of events, modify them with a single gesture, and visualize the result in some non-numeric way. But to keep my bearings I also have to keep the written score at hand, and I get dizzy glancing back and forth between the two things.

Perhaps the ideal is to integrate some tools like this with the notation view. I can’t quite picture how it would be done, but fortunately the MS developers are not limited by my imagination.

Finally, are you familiar with BSG’s DockArticulate plugin for MS? It provides an Inspector-like access to note timing.

In reply to by scott.1

I'm getting the sense MS 4 is heading towards what you describe in the second para, and it's what I hope to see. And boy, do I understand the "dizzy" comment. Nausea sets in after an hour or so of flipping between score and PRE.

I didn't know about DockArticulate. Thanks for bringing it to my attention. I'll check it out immediately.

sounding great, I am used to accept the playback in my Sibelius and send to the people who is gonna perform it, just to get an idea of how the performance will sound like. But this is way better, although I had a hard time hearing the brass, also, my experience is that brass is hard get to sound authentic, any thoughts on that?

In reply to by stefantrumpet1

The piece uses very little brass except for the horn quartet, so there's not much to hear. I agree brass is hard to get to sound authentic, at least in orchestral music. Most brass samples have too much character and stand out like soloists, when what's needed is something that blends with the orchestral fabric and only occasionally has a solo-ish role.

Do you still have an unanswered question? Please log in first to post your question.