What MusicXML actually is
A sheet music PDF is a picture. It knows where the ink is, and nothing about the notes. MusicXML is the opposite: a text file that records every pitch, duration, clef, key, tie and dynamic as data, which is why a notation program can open it, transpose it, re-engrave it and play it. Converting a PDF to MusicXML means reading the picture and writing the music out as data, and that job is called optical music recognition.

- A PDF of sheet music contains no notes, only an image of them. MusicXML contains the notes.
- Sheetly imports a PDF, reads it with optical music recognition, and exports MusicXML that opens in MuseScore, Sibelius, Dorico and Finale.
- Pitches and rhythms convert well from a clean printed PDF. Lyrics, fingering and expressive text are the parts to check by hand.
- If you intend to correct a long score note-by-note, a desktop tool with an editor is the better purchase. Sheetly does not have one.
MusicXML is the interchange format the whole notation world agrees on. MuseScore, Sibelius, Dorico, Finale, Guitar Pro and most practice apps can read it, and most can write it. That is why it is the format worth asking for: a MIDI file carries pitches and timing but throws away the notation, while MusicXML keeps the clefs, the key signature, the beaming, the ties and the two staves of a piano part as two staves. The difference between the two is explained in full here.
The catch is that a PDF almost never contains MusicXML already. Even a PDF exported straight from a notation program is usually just vector drawing instructions with no musical meaning attached. So the conversion has to be done by recognition, the same way you would scan a photographed page, and the same cautions apply.
PDF to MusicXML in five steps
The whole conversion takes about as long as making a coffee, and most of that is the recognition running. Sheetly treats a PDF exactly like a photographed page, except that the input is already flat, square and evenly lit, which is the best case a scanner ever sees.
- Import the PDFOpen Sheetly, choose import rather than the camera, and pick the PDF from Files, iCloud Drive, Mail or wherever it lives. Multi-page PDFs import as one piece; you do not need to split them first.
- Let the recognition runThe pages are sent to our servers, read, and handed back as a score. A few seconds per page is normal. You need a connection for this step, and for nothing afterwards.
- Press play, once, all the way throughThis is the cheapest check there is. A misread clef, a missed key signature or a dropped bar announces itself in the ear long before you would spot it on screen. Note the bar numbers of anything that sounds wrong.
- Export MusicXMLFrom the score, choose export and pick MusicXML. Sheetly also exports MIDI and PDF; for anything you plan to edit, MusicXML is the one you want. Send it to Files, AirDrop it to a Mac, or share it straight into your notation app.
- Open it and look at the first barIn MuseScore, Sibelius, Dorico or Finale, check the clefs, the key signature and the time signature in bar one. If those three are right, the rest of the file is usually right with it. If one is wrong, fix it there and most of the score falls into place.
That is the whole flow. The rest of this page is about what the file will and will not contain, because knowing that up front saves a lot of hunting later.
What survives, and what to check
From a clean, printed PDF, pitches, rhythms, clefs, key and time signatures, ties, and the grouping of a piano part into two staves all convert well. Text, fingering, dynamics and lyrics are where every recognition system, ours included, is least reliable, and they are the items worth a deliberate pass before you trust the file.
What usually comes through cleanly
- Pitches and rhythms. The core of the music. On Sheetly’s own 31-piece benchmark of clean printed scores, measured by strict F1, where a note only counts when pitch, onset and duration are all correct, the score was 96%. A clean PDF is exactly that kind of input.
- Clefs and key signatures. Including changes mid-piece, which is where a missed one does the most damage.
- Time signatures and bar structure. Bars that do not add up are the most common sign something went wrong, and the easiest to see in a notation editor.
- Ties and dots. A tie and a slur look almost identical on the page; the distinction matters for playback and is usually right but worth a glance in long held passages.
- Staff grouping. Two staves braced as a piano stay a piano. Four staves of a string quartet stay four parts. Why that is harder than it sounds is here.
What to check by hand
- Voices. A staff with two voices, stems up and stems down, is the hardest thing to read and the first thing to check if the rhythm of a bar sounds crowded or thin.
- Lyrics and text. Tempo marks, expression text, rehearsal letters and lyrics may be missing or misplaced. Re-entering them in the editor is quicker than hunting for them.
- Fingering and articulation. Treat these as absent and add the ones you need.
- Repeats and endings. Repeat barlines and first-and-second-time brackets are read, but a missed one silently shortens the piece. Count the bars against the PDF if the length feels wrong.
- Tuplets. Triplets and other tuplets are a known weak point across optical music recognition generally. A bar that will not add up is very often a tuplet that lost its bracket.
Many PDFs in the wild are scans of paper: slightly skewed, a little grey, with the spine shadow of a library book down one side. They are still far better input than a phone photograph, but they behave like scans, not like exports. On a public suite of 60 systems cut from real scanned pages that we did not author, Sheetly scored 90% mean and 99% median strict F1, and produced usable output for all 60. The median tells you most conversions are nearly perfect; the mean tells you a few are not. Listen before you trust.
When a desktop tool is the better choice
If your real job is to produce a corrected, publication-quality MusicXML file from a long or complex score, a desktop recogniser with a built-in correction editor is a better tool than any phone app, including Sheetly. ScanScore, PhotoScore and the free, open-source Audiveris all let you fix misread notes in place before exporting, and Sheetly does not.
That is an honest limitation, and the right tool depends on what you are converting and why:
- ScanScore is built around exactly this workflow on a desktop: scan, see what it got wrong, correct it, then export MusicXML. It is paid, tiered by how many staves it will read, and its trial limits MusicXML export to a few measures, so plan an evaluation around that.
- PhotoScore & NotateMe is the long-standing professional option, closely tied to Sibelius, with a mature correction workflow. It is priced for professionals and expects a desktop.
- Audiveris is free and open source, expects flat scans, wants some setup, and includes an editor. On our clean benchmark it scored 88% against Sheetly’s 96%. For a clean PDF on a desktop computer it is still a genuinely good answer, and you should try it before paying anyone.
Where Sheetly is the better choice is the case most people are actually in: you have a PDF on your phone, you want to hear it, practise it and maybe hand a MusicXML file to a notation program for light editing, and you would rather not sit at a desk to do it. For that, the phone is the right tool and the correction happens in whatever editor you open the file in.
We have not measured the closed products, ScanScore, PhotoScore or Sheet Music Scanner, through our own harness, so the numbers above compare Sheetly only with the open-source systems anyone can run. Their absence is a gap in our testing, not a verdict.
Before you export, ask about it
A MusicXML file is where most conversions end. In Sheetly it is one of the things you can do with a score that has been read, and not the most interesting one. Once the PDF is a score, you can loop a passage, slow it down, play it on piano, organ, flute or another instrument, record your own take against it, and ask the coach about what is written on the page.
For a conversion job specifically, the coach is useful as a second pair of eyes. It has the actual notes of the score, so it can tell you what time signature it sees, how many bars it counted, or whether bar 17 really has a triplet in it.
Ask Sheetly: a coach that has read your score
Not a chatbot bolted onto a music app. The coach has the actual notes of the page you just photographed, so you can ask about your bar 12 and get an answer about what is written in it.
We know of no other sheet-music scanner that does this. Every other tool on this page hands you a transcription and stops.
Questions
Can I convert a PDF to MusicXML for free?
Yes, with a desktop computer. Audiveris is free and open source, expects a flat scan or a clean PDF, and includes an editor for corrections. Sheetly is free to download on iPhone and iPad, and exporting MusicXML is part of its Pro subscription.
Does the PDF have to be a clean export, or can it be a scan?
Both work, and a clean export is the easier case. A scanned PDF behaves like a scan: skew, grey backgrounds and spine shadows lower the accuracy somewhat. On a public suite of 60 systems cut from real scanned pages, Sheetly scored 90% mean and 99% median strict F1, so most scanned PDFs convert nearly perfectly and a few do not.
Will the MusicXML open in MuseScore, Sibelius, Dorico and Finale?
Yes. MusicXML is the interchange format all four read. Sheetly exports standard MusicXML, so the file opens in any of them, and in most other notation and practice programs too.
Can I convert a multi-page PDF to MusicXML in one go?
Yes. A multi-page PDF imports into Sheetly as one piece, the pages are read in order, and the MusicXML export contains the whole score. You do not need to split the PDF first.
Why does a bar in my MusicXML not add up?
A bar whose durations do not fill the time signature is almost always a tuplet that lost its bracket, or two voices that were merged into one. Both are known weak points of optical music recognition in general, and both are quick to fix in a notation editor once you know which bar to look at. Pressing play in Sheetly first is the fastest way to find that bar.