Importing PDF files?

• Dec 21, 2021 - 15:44

Just wondering if any progress has been made to import PDF files?


And if neither the automated service nor running Audiversis yourself produces results - that would be normal if the original is handwritten or poorly scanned - then you can always try any of the various commercially-available music recognition apps. They all produced MusicXML that MuseScore can then import.

Well, long story short, dowloaded open source from github and compiled audiveris for my MacOS.

Being a retired software engineer and familliar with the tools, good news was able to compile successfully. But had to install some other packages to accomplish this. Notably, older version of java (jdk11) and freetype. best to use Brew.

Bad news, importing a .pdf to generate a .xml fails with java exception out of Memory heap. Posted the issue on github. we'll see if i can get this resolved.

The import which sends to the audiversi server definitely doesn't work. keep getting unsuccessful with any pdf.

I will try to export .mscz file to .pdf and try that to see if there's an issue with either webserver or the opensource version.

If anyone trying this, you really need some software development skills to get through it.

Any other suggestions for apps that have been successful i'd like to hear

OK so I exported a musescore file to a pdf.

Imported the pdf into audivertis opensource locally and failed. also tried the Audivertis web server again with the pdf and got the unsuccessful message.

So bottom line unless i hear back from gitHub team I'm moving on and look for a different application

In reply to by drummerMan

As regards commercial OMR software I have successfully used:
a) SharpEye: very old-fashioned in look and feel but excellent at note recognition (not so good at lyrics OCR); cannot use PDF file as input (I use TIFF in preference to BMP).
b) PhotoScore: more up-to-date, better UI, can use PDF file as input.

I have also heard good reports of ScanScore, but I have not yet used it.

In reply to by drummerMan

"SharpEye actually scans and reads the TIFF"
I don't want to create a misunderstanding here. So let's be clear: I have never tested SharpEye with a multi-page TIFF file. I always use individual TIFF files, one per page of the original document.

And I find that black-and-white mode works better than greyscale. In GIMP terms, I always convert with Image > Mode > Indexed... > Use black and white (1-bit) palette.

In reply to by drummerMan

Earlier you wrote:
Any other suggestions for apps that have been successful i'd like to hear
OK so I exported a musescore file to a pdf.

There is a difference in whether the music was exported as a pdf file from a score editor, or whether the pdf was created from an image (e.g., scanned to pdf).


For a music pdf exported from a score editor, see
Available for MacOS, and a free trial, too.

In reply to by drummerMan

You really don't need to find out how any music pdfs were generated. The conversion results will speak for themselves. Some conversion software work better with one. Some can handle both.
Years ago I used Audiveris - before its current use by MuseScore - and recall scanning printed music (so generating a "picture" of the music). I had fairly decent results with lead sheets.

Anyway (as shown in my post above), you can always "click and scrub" on a pdf score to see whether it is a picture, or whether "scrubbing" reveals those discrete elements as produced by a score writer. Then choose a converter to use.

In reply to by Jm6stringer

Well that was easy and troubling. Using pdftomusic demo version only converts the first page. Musescore opens up with either the file dumped as text or error msg indicated file is corrupt.

So not sure if it's because has not been fully exported.

I opened a musescore file (.mscz) that was 1 page and exported as a pdf. Opened in pdftomusic and exported to .xml. Musescore did pretty good job of bringing it in.

So bottom line is that is it possible that pdftomusic creating only one page of multipage is not a valid .xml file for Musescore?

Other considerations...

I don't know how savvy most OMR's (optical music recognition) are when it comes to drum notation, which "looks" different from standard musical notation for other instruments. Related to this is Tablature (e.g., for guitar) which can also be problematic depending on the OMR's ability.

Also, for actual lyrics "recognition", see:

In reply to by Jm6stringer

Thought I share my experience dealing with Myriad - pdfToMusic app.

Downloaded the macOs trial version. Although exported to xml, Musescore didn't read the file correctly. Got a measure followed by numerous empty measures, another measure followed but empty measures, etc... see attached files 4-6.

Then I tried the Windows trial version. but, they purposely disabled exporting xml. Why? They said because that's why it;s a trial version.. unreal baloney...

So I was able to get a forum response from i guess the admin to read the pdf file and export to xml. He shared the xml file with me to test. Musescore opened the file, and got an error message indicated the file was corrupt. I disregarded the message anyway to see if it could still be read. It sure did, but not very readable. see attached files 1-3

So bottom line although xml is a standard, is it possible Musescore is not interpreting the xml file correctly or is it Myriad not abiding to xml?

Musescore is running on my Imacs and powerbook all up to date MacOS.

In reply to by drummerMan

The corruption messages mean that the scanning software wasn't able to correctly understand the notated rhythms and put too many or too few beats in a measure. You'd need to fix those as described in

It's pretty normal than these programs will not be able to sort everything out correctly, and it will take considerable effort to fix. I thought I posted a response here quite a while ago, but apparently I didn't. What I meant to post went something like this:

If you're interested in checking out what's possible as an intellectual exercise, go ahead, it can definitely be interesting.

But if the goal is to actually get music into MuseScore, it's usually much simpler to just enter it normally. AI technology just hasn't progressed to the point where computer programs can do a reliable job of turning pictures of scores into actual scores. Some do better than others, but normally a ton of work is required to fix the errors and corruptions, and that work is usually far more difficult and requires far more program expertise than simply entering the notes normally.

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