MIDI import and MuseScore´s interpretation

• Mar 10, 2014 - 12:08

Hello,

A very basic qestion concerning the MuseScore notation engine when
I import the MIDI information. For me the biggest problem seems to be that
MuseSocore´s interpretation of the note picture is too complicated, or
at least MuseScore makes the note picture very "messy". Let me give
an example: I play a very simple one-hand piano tune with my sequencer,
export the MIDI file and open that in MuseSocre. MuseScore´s interpretation
how to trascribe the note picture, i.e. note lenghts, legatos, rests etc.
is very "busy", see the file example attached. Of course it is theoretically "right" but
MuseScore has made this very simple melody picture to look very complicated
and hard to play for someone who gets the score.
Is there some easy way to "simplify" the way MuseScore interprets the
MIDI information? Of course I can always edit manually the whole
score but in this way I loose the whole idea of importing MIDI files.

Harry

Attachment Size
MuseScore MIDI test.mscz 3.33 KB

Comments

Try with a recent nightly build, quite a few improvements to MIDI import have been done there.
Then please note that MIDI is just not designed to carry all the information a scoring app would need...

In reply to by Jojo-Schmitz

Thank you for commenting. What is "a recent nightly build"?
My problem is more like opposite you refer: the MIDI file I import to the MuseScore carries TOO MUCH information, i.e. MuseScore intrepretates also all unnecessary details I don´t want to see. How could the notation programm know what is "unnecessary"?
I think that that solution towards I´m looking for is a rather simply algorithm (I imagine) inside the MuseScore, which converts the MIDI information in an adjustable "roughness" or "coarseness" to the note picture. E.g. if I play a piano tune which includes 1/4 notes and 1/8, which sounds perfect, but some 1/4 notes are mathematically few nanoseconds longer and some 1/4 notes few nanoseconds shorter, and sometimes even I´ve let some note "ring" a bit longer wihtout meaning this should show in the note picture, MuseScore faithfully reproduces the "mathematical truthness" which can be really messy when printed.
I should be able to "tell" the MuseScore, that I want to convert the MIDI file in certain roughness level, even adjust the tolerances of each note and also each rest.
Quantitization notes inside the source (in my case inside the sequencer) is not the answer because this often destroys the feeling and the nuances of the music and doesn´t necessary correct the note lenghts but only the positions. Harry

Welcome aboard.

The forum pages are rife with issues regarding midi import. (You can do a search in the search box.)
Advice ranges from: 'Just think about how good a sight reader you'll become!'... to... 'Quantize'.

Basically, midi stores note on/off events for a machine, whereas music notation is geared toward human performers - so while a machine may be flummoxed by the vague instruction 'ad lib', it will get back at us humans by employing the 'redundant and/or unnecessary ties' technique - messy, but theoretically 'right', as you mentioned. :-)

The good news is that the 2.0 version of MuseScore, currently under development has made improvements to the rendering of midi importation.
See:
http://musescore.org/en/handbook/comparison-stable-prerelease-and-night…
http://musescore.org/en/download#Nightly-versions
where you can test drive the new version via a nightly build.

Still, when all is said and done, sometimes a midi sequencer can help to rigorously quantize a score to make the transition to notation less 'busy'.
In the past, I have used Music Creator 5 for displaying midi files as music notation with better readability.

Regards.

In reply to by Jm6stringer

Hi, thank you for commenting. I have to check the Music Creator 5. See also my
comment above concerning the quantitization. I look forward to the MuseScore 2,
I´m not a programmer, but I could imagine that adding the algorithm, which lets you adjust the roughness of the conversion (tolerancies), shouldn´t be too complicated to make, if the developer just saw the importance of creating notation picture of the MIDI files (which I need in larger scale). I used make music with the Logic sequencer which included a rather good notation feature besides the typical MIDI piano roll, there the note picture was pretty clear compared to the one you get when you import MIDI file to the separate notation program. Unfortunately the Live, which I use at the moment, don´t include any kind of notation feature. Harry

In reply to by Harry_

Sequencers and music notation apps...striving for 100% congruity. Is it possible?

You mention that quantization 'often destroys the feeling and the nuances of the music', which is why some sequencers have something similar to a 'human playback' setting to offset rigorously quantized music. The nanosecond variations in performance you mention is what distinguishes a human from a machine.
If one composes via midi recording using a sequencer, the timing variations of human performance are duly recorded to produce a faithful replica. Meter, key/time signatures, notated durations (eg.half note, quarter) mean less to a sequencer than does performance playback.

MuseScore is primarily a music notation application. It produces scores to be performed by humans. As such, things like time signatures and note durations help the musician realize the composer's musical thought.
Regarding 'feeling' or 'nuance' - consider the metronome (a sort of 'quantizer') as used by a beginner: Once a piece is mastered, the metronome is often set aside, as it would actually interfere with 'feeling' and 'nuance'. For example, even though 4 sixteenth notes may be duly notated, there might be an interpretative pause added which technically won't add up to *exactly* a full beat. So, music notation is not as rigorous for human interpretation as midi must be for machine interpretation.

Concerning playback, though, MuseScore does a decent job. In past iterations, the only instrument available was piano - basically as a sort of 'spell check'. With the newer versions, people on these forums have been pushing the envelope - actually using MS as both a notation and a performance tool - tweaking soundfonts and note properties (such as ontime/offtime) to create 'lifelike' music.

Regards.

In reply to by Jm6stringer

Well, I believe that all (above discussed) depends on what you are looking for.
My intention is to get some simple musical ideas written in sequencer in to notes, to make possible to someone else to read/play/arrange the basic themes/harmonies. Someone else may
use the notation program differently.
The note picture should be as simple as possible, rather over-simplified than
including every single state-of-the art detail. Clarity is here everything to me.
I was hoping that I could use MIDI import to save time - because every note is already written/played.
So far writing everything from scratch by using the old-fashioned paper-and pen has
been the quickest way to get want I want. I was hoping that MuseScore could help me
- computers and software do make today a lot of the more complex things.
Harry

In reply to by Harry_

...though there is a tradeoff using sequencers.
If you want the 'note picture' to be as simple as possible (here I'm assuming you mean the written score), then 'feeling' and 'nuances' have to be sacrificed when entering notes in a sequencer. Rigorous quantizing facilitates better midi importation into a score writer in order to produce music notation.

Having said that, consider this:
If you wish to 'get some simple musical ideas written into notes', you can use a midi keyboard for note entry directly into MuseScore.
http://musescore.org/en/handbook/note-entry#MIDI-keyboard

Regards.

In reply to by Harry_

I think your expectations are not reasonable. There are two main problems here.

1) A MIDI file contains too little information - and it is indeed too little, not too much - as discussed. When you play a note, all the MIDI file says is when it started and when it stopped. It doesn't say whether it should be represented as a quarter note or eighth note, it doesn't say whether it belongs to voice 1 or voice 2 and should thus be drawn with stem up or stem down, it doesn't even say which hand played or which staff it should be drawn on. MuseScore has to *guess* answers for all of these. Bad guesses = ugly score.

2) Not everything that can be played on a keyboard is possible to notate cleanly in any objectively correct way. Overlapping notes - where you press one note before releasing the previous one - means multiple voices, and while most notated music is done using only 2 voices for clarity, it's likely you are doing more overlaps than that (eg, arpeggiating a chord and holding each note down as you do so). The way a human would have notated that is without the overlaps, expecting a human player might overlap the chord himself if he felt like it. Here again, MuseScore would have to *guess* at how much overlap to throw out.

Any score that *accurately* captures what you play is going to be ugly. Conversely, any score that looks good is going to be a very poor approximation for what you actually played, full of lies like removing overlaps. At best you can maybe hope that by some miracle, the particular lies told by the software will be reasonable lies, but this is entering the are of artificial intelligence and is nowhere near as simple as merely "quantizing" note values.

In reply to by Marc Sabatella

Most informative reply, thank you.

"At best you can maybe hope that by some miracle..."
If you know what your are talking about in the "bit-level" (and I don´t have any reason to doubt that), the situation from my point of view/the way I planned to use the MuseScore, is
much worse than I though.

I believed you don´t need "artificial intelligence" (in the demanding sense, in the broader sense most softwares are "artificial intelligence", I guess) to make the rules that cover the most frequently used cases, and in this way eliminate most of the "noise" you can see e.g. in the small sample I´ve attached.

A couple of things:
When talking about musical "nuances", I would separate the played music and the
notating. The details blow life in to the music (therefore e.g. quantization in something I rarely use) but in the notation I accept a "rough estimation" or the "bone" of the performance, in the note picture the clarity and easiness to read is the one I prefer, even at the exepence of some accuracy what comes to e.g. legato, rest etc. characters). I don´t WANT all the details. I underline that this is the way I, in the most of the cases, use the notation. I think the best notation software is flexible, one where you can adjust the amount of the details depending on the purpose.
An other point: My workflow includes the use of the sequencer because I create the arrangement and test different sound alternatives (both digital and analog) all the time, these are "tailored" inside the composing process, therefore writing music with the notation program connected to the midi-keyboard is in my case rarely used. H.

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