For piano teachers

Sheet music OCR: what music recognition does, and how to get a usable result

"Sheet music OCR" is what most people call optical music recognition, or OMR: software that looks at a picture of printed music and works out the notes. It's the step that turns a photo or PDF into something you can play back, edit, transpose or export.

This guide explains what OMR does, why it goes wrong, how to check the result, and how to choose a tool. We make PlayableNotes, a music-recognition tool for piano teachers, so we'll say where it fits and where another tool fits better.

Try music recognition on your own page →

50 free reading credits for new accounts. No card needed, no subscription.

How is music OCR different from text OCR?

Text OCR reads characters in a line, one after another. Music is harder, because the meaning of a symbol depends on everything around it:

• Position is pitch. A notehead's height on the staff decides the note, and the clef and key signature change what that height means.
• Things happen at once. Chords, two voices on one staff and two hands on two staves all sound together and have to line up in time.
• Symbols reach across the page. Ties, slurs, beams, repeats and first and second endings connect things that aren't next to each other.
• Small marks matter. A dot, a tiny accidental or a natural sign changes what you hear.

So an OMR tool doesn't only find symbols. It has to rebuild the music's structure: which notes belong to which hand, how each bar adds up, and what order the repeats play in.

What comes out of music recognition

Once the notes are recognized, you can usually:

• Play it back to hear the piece.
• Edit it, to fix mistakes or change the music.
• Export MusicXML, the standard format for notation software such as MuseScore, Sibelius and Dorico. See image to MusicXML.
• Export MIDI, a performance format for DAWs, digital pianos and practice tracks. See sheet music OCR to MIDI.
• Print a clean PDF or share the score.

OMR isn't the same as audio transcription, which starts from a recording and guesses the notes from sound. If you have the sheet music, OMR is the more direct route.

Why sheet music OCR makes mistakes

No recognition tool, ours included, gets every page right. The usual causes:

The picture. Curved pages near the spine, shadows, glare, blur, a cropped edge or a low-resolution scan. This is the cause you can control most easily.

The engraving. Old, faded or crowded editions, very small print, and photocopies of photocopies.

The notation. Cross-staff passages, several voices on one staff, tuplets, grace notes, ornaments and unusual symbols.

Handwriting. Most OMR is built for printed music. The open-source Audiveris project says plainly that handwritten scores aren't supported. Klangio's help center says Scan2Notes is trained on printed music and results from handwriting "can vary extremely".

What the tool chooses to read. Tools differ in which markings they pick up. Scan2Notes' help center, for example, lists dynamics, articulations and grace notes as currently not supported. Check what your tool reads before you assume a marking will come through.

Even the developers say so. The Audiveris handbook calls its engine's accuracy "still far from perfection" and builds its interface around checking and correcting. That's the right attitude for any tool.

How to get a better result

Before you read the page

1. Photograph it flat, sharp and evenly lit, with the whole page in the frame. One page per photo.
2. Use the cleanest copy you have. A publisher PDF or a clean public-domain edition beats a photo of a worn book.
3. Don't over-process the image. PlayScore's own tips warn against document-scanning apps, which can degrade the image. Upload a plain photo first.

After it's read

1. Listen while following the original. Errors are easier to hear than to see.
2. Check the usual suspects: accidentals, ties, ledger-line notes, bar lengths, repeats and endings, clef and key changes.
3. Fix it at the source. Correct notes in the recognition tool before you export, so every format you download is right.
4. Check again in the destination program. No format carries everything. In our PDF to MuseScore worked example, all 333 notes and both repeats came through. The tempo word "Spiritoso" and some expression details didn't, and were fixed in MuseScore.

Choosing a sheet music OCR tool

The right tool depends on what you need to do after the page is read, more than on recognition alone. A short summary from each tool's own site (checked 16 October 2026):

For prices, limits and a longer side-by-side, see the best sheet music scanners for piano and Scan2Notes, PlayScore 2 and ScanScore alternatives.

A fair test: take the same page, read it in each tool you're considering, and count how many corrections you need before it sounds and looks right in the place you'll actually use it.

Music recognition in PlayableNotes

PlayableNotes is built for one job: getting a piano page to a student-ready version.

1. Snap it. Upload a photo (one page per photo), a screenshot or a PDF. You check the pages and their order, and the credit count, before anything is read. Reading usually takes a minute or two.

2. Check it. Press play and the notes light up as they sound on a sampled piano. Click any wrong note to fix it, ask the AI to fix a passage, or retake the photo. A reading that fails costs nothing.

3. Use it. Download MusicXML, MIDI or a PDF in Letter or A4. Or type what your student needs ("easier left hand", "no octaves", "put it in C"). Each change usually takes about a minute, so a first student version takes a few minutes. Send a free practice link the student opens without an account.

The Minuet worked example shows the whole process on a real page. It includes the credit quote, the check and one note restored by hand in the arrangement. If you want to hear a page right away, start with scanning piano sheet music. For a notation file, see converting a piano PDF to an editable score.

What it costs: reading uses 1 credit per line of music (both staves together). New accounts get 50 free credits. Packs cost $10 for 100, $25 for 300 and $70 for 1,000, credits never expire, and there's no subscription. Playback, editing, arranging, exports and practice links are free. Credits are sold to US billing addresses only, for now. See pricing →

Read your first page free →

Upload a page. Your first 50 lines are free.

Frequently asked questions

What is sheet music OCR?

It's software that reads a picture of printed music and turns it into notes a computer can play, edit and export. The proper name is optical music recognition (OMR).

Is there free sheet music OCR?

Yes, with different limits. Audiveris is free, open-source desktop software for printed music. PlayScore 2's free app plays pages from the camera. Scan2Notes offers a free preview of the first 12 measures. PlayableNotes gives new accounts 50 free reading credits, enough for roughly 7 to 10 typical piano pages.

Can sheet music OCR create MIDI?

Most tools can export MIDI after recognition. In PlayableNotes, MIDI downloads are free and the right and left hands come out on separate tracks.

Can it read handwritten music?

Usually not reliably. Audiveris doesn't support handwriting, and Scan2Notes says results from handwriting vary. PlayableNotes doesn't advertise handwriting support. Try a page before relying on it.

Which sheet music OCR is the most accurate?

We haven't tested these tools against each other on the same scores, so we don't claim a winner. Results depend heavily on the page. Try your own music in each tool and count the corrections.

Does my music leave my device?

PlayableNotes' terms say it processes your music with AI providers that read the music, make the changes you ask for and check uploaded photos (OpenAI and Cloudflare). Your uploads stay private to your account.

How long does it take in PlayableNotes?

Usually a few minutes: reading takes a minute or two, and each change takes about a minute.

Does PlayableNotes read music for other instruments?

No. It reads piano music only: a treble and a bass staff played together.