Problems with score cleaning/editing (automatically-generated mscz file)
Hello,
I was given a MuseScore file that had been automatically extracted from an IMPSL score (Schubert Ave Maria).
That file has problems, and I am looking for the simplest ways to correct for those
Among the (many) problems I would like to correct
- Consider the left hand of the piano (mscz file), it seems that several layers are involved and there are therefore systematic spurious indications: rather that rewriting from scratch, is there a simpler way to correct it ? (e.g. force-combine layers into one ?); how to do that ?
- I reasoned that if several layers are involved for that left-hand part, I could use a simple trick of copying that part, paste to a new file, and then convert that file to midi. Then open that midi file with MuseScore.
But this fails miserably (see attachment), the midi export is apparently not respecting properly the note length. - there are also greyed "plus signs" that are scattered: they appear to be not-editable; to what do they correspond / how to get rid of them ?
More generally, I am dreaming to be a bit more efficient when working with automatically generated scores (that are often a bit weird in their structure): how to efficiently tap into such automatically-generated data as a boostrap to get the final correct file... (or is it overall more time efficient to rewrite from scratch...?)
Thanks for your advices !
| Attachment | Size |
|---|---|
| Schubert, Ave Maria; Score problems.mscz | 21.28 KB |
| Ave Maria; exported midi from left-hand piano.mid | 930 bytes |
| IMSLP284003-PMLP16143-Nuptial_Ave_Maria Si bémol Maj Original.pdf | 765.68 KB |
Comments
And on the piano right-hand, how to painlessly get the sextolet structure ?
In reply to And on the piano right-hand,… by francois71828
I'm a great supporter of optical music recognition (OMR) and use it frequently.
But it will be difficult or impossible for software to succeed with the PDF you attached. A human musician will recognize that the sextuplets shown in the first measure continue as sextuplets for the following measures. But without explicit tuplet markings, no OMR software can be relied on to have the same insight as a human musician.
I think you would find it easier to notate the first measure as empty sextuplets, highlight the measure and repeat-copy the empty sextuplets with R until the end of the piece. Then add the notes manually: in the end it's likely to be quicker than struggling to correct the painful and broken efforts of an OMR import.
In reply to I'm a great supporter of… by DanielR
I am interested to learn more about basic tricks to make musical OCR be useful. (I stumbled repeatedly in problems and worthless output).
In the case example (even though I just got the info that the mscz file was available): for a better OCR, should I have tried to add sextuplet notations on the PDF before OCR ? (say using Inkscape one page of the pdf at a time) ?
In reply to I'm a great supporter of… by DanielR
My reasoning is that the midi does not care that much about the time signature, eg. 2 triplets or 1 sextuplet is the same at the midi level, except for the meta-message that specifies it: are there some free midi editor available that would allow for brute-force change of the time signature, without altering the midi data itself ?
In reply to And are there ways to… by francois71828
Definitions, for clarity:
OCR = optical character recognition (text)
OMR = optical music recognition (music staves with notes)
"...for a better OCR, should I have tried to add sextuplet notations on the PDF before OCR ? (say using Inkscape one page of the pdf at a time) ?"
I do sometimes use a graphics editor to add missing information to the page of music, either before I try using OMR or if the first OMR attempt fails completely.
"And are there ways to somehow get rid of the strong "bar-formatting" of musescore, by using midi-export /midi import ?"
I am not an expert in how the various OMR systems work, but they must all rely on recognising and observing the current time signature. The OMR software will use the expected number of beats in a measure to guess at note duration where this is unclear because of printing faults. For example: a half note (minim) where the notehead is wrongly filled in, or a quarter note (crotchet) where there is a white gap in the notehead.
"... are there some free midi editor available that would allow for brute-force change of the time signature, without altering the midi data itself ?"
I don't use any MIDI editors, so perhaps someone else can answer this one.
However this approach seems to me to be a blind alley: if the basic time signature isn't clear without prior editing, any OMR attempt is going to fail anyway.
FWIW, on a common score like this, it's far more efficient to just download one of the very many excellent versions already entered on musescore.conm, like this one: https://musescore.com/openscore-lieder-corpus/schubert-franz-op-52-no-6…
In reply to FWIW, on a common score like… by Marc Sabatella
OK, great !
(I had searched the MuseScore website with:
ave maria Schubert,
and did not find it.
In reply to OK, great ! (I had searched… by francois71828
Which website, musescore.com or musescore.org?
The former has several hits: https://musescore.com/sheetmusic?text=ave%20maria%20Schubert
The latter is not supposed to ;-) (except via the "Search for shee music" files in the header, which I used for above result)