Last week’s discussion of software to convert PDF examples from public domain sites into editable documents elicited a reaction from one software group eager to be recognized:
MuseScore is a free and open-source notation program for Windows, Mac and Linux. The software is an alternative to the big two of Sibelius and Finale, and users of those programs will find MuseScore to be familiar. As with other open-source programs, it’s not a straight transition from the programs you know and love to a “work in progress” that’s constantly under development, and open-source software usually operates with less of a safety net in terms of documentation and support than commercial products. That said, there are certainly advantages to open-source software, none more obvious than that “free” word.
I left MuseScore out of our initial conversations because I wasn’t aware that they had a feature to import PDFs into editable files… they call it an “experimental” feature (link behind login- free MuseScore account required). It is, however, a valid option for processing files into MuseScore from public domain sources as .PDF. So how does it work? In their demo, MuseScore sent two files: the original PDF downloaded from CPDL (Per Avem, ed. Marco Croci), and the processed MuseScore output file (they present it as “the result with zero clean up”).
(original)
(processed into MuseScore)
After comparing them side-by-side, a few things become apparent. First, there is a very high degree of accuracy with regards to the notes. The SAT lines are accurate, and two notes are missing from the Bass, easily re-added in editing. Certainly had the original been a handwritten manuscript, there would be more issues with pitch accuracy, but analyzing a printed score is fairly reliable at this point. Second, the text is a mess in the output. This is pretty standard for many OMR programs, and it’s just hard to accurately detect which notes the text should be attached to. I often find that this is less of a burden than you might think: if I’m importing a file back into a notation program, I’m usually doing it to produce a recording (which will not require the text) or to re-arrange or re-structure the piece, in which case I’ll have to revise the text setting anyways.
The more suspect areas of OMR processing usually involve rhythm, and here we do see some issues that will require cleanup:
(edited MS file, ex. 1)
Do you see the issue? Hint: it’s not the pp high F. The scan has inserted eighth-rests behind some of the printed notes. Easy to delete, but they should be removed in editing.
(edited MS file, ex. 2)
This is a little more obtuse.
Aside from rhythms, the dynamics and articulations are the next categories of markings most likely to have issues, and here there are some errors as well. As with text, asking a piece of software to make an accurate determination of where dynamics are supposed to be attached, as well as assigning floating markings such as articulations, steps far outside the comfort level of computer processing. Again, how much editing this requires will largely be determined by what you want to do with the finished file (and how erroneous the markings are).
The examples that MuseScore provided are great demonstrations of both the capacity and limits of OMR software designed to convert scanned or downloaded .PDF files into editable notation. Where even until recently, reasonable minds differed on whether OMR software saved time over re-inputting a score from scratch, I believe that these two scores side-by-side show that processing and editing a digital file downloaded from a public domain source is a relatively easy process. In addition, now the MuseScore has this capacity, this may be a viable option for you if you did not previously own one of the commercial programs we discussed last week.
Gary Rader says