OMR sheet music recognition (image to MusicXML)

What is this?

OMR stands for Optical Music Recognition — the automatic recognition of musical notation from images. With the OMR feature you upload a scanned or photographed score, and Chorilo turns it into a MusicXML file.

This is useful when:

• You only have paper scores and want to use them digitally
• You want to use the voice trainer, which prefers clean MusicXML as the best source
• You want to transpose a piece or extract individual voices
• You need an editable version of the score (for example for notation software)

How to use OMR

1. Open the Sheet music area and click Upload sheet music.
2. Choose OMR mode instead of the regular upload.
3. Upload the image — PDF, PNG or JPG with the score pages.
4. Enter title and composer as you would for a regular upload.
5. Start the recognition. Processing takes between 1 and 10 minutes depending on page count.
6. Check the result and correct any errors manually.
7. Save the finished MusicXML as a new piece.

What OMR can and cannot do

Works well for:

• Clearly printed, modern editions
• Straight scans with good contrast
• Standard notation (no experimental notation)
• Resolutions of 300 dpi and above
• Black and white or grayscale

Has trouble with:

• Handwritten scores (often only approximate recognition)
• Skewed or wavy originals
• Note heads that are too small or too large
• Complex notation (such as polyphony on a single staff)
• Poor contrast or stains
• Very old or faded prints

Permission

To use OMR you need the sheet_music.create permission. Processing may incur a fee depending on your plan — you see the cost before you start.

What happens after recognition

After processing you get:

• A MusicXML file stored as a piece in the library
• A side-by-side comparison between original image and recognized result
• A correction tool for obvious errors (wrong notes, rests, accidentals)
• An audio preview of the recognized notes

You can edit the MusicXML further at any time, transpose it, split it into voices or use it as a source in the voice trainer.

Tips

• Scan straight and with high contrast — preferably as a PDF at 300 dpi or higher.
• Crop the images before uploading so only the staff is visible (no margins, no annotations).
• For multiple voices on one staff it pays to separate the voices beforehand — recognition becomes more accurate.
• Always review the result thoroughly before releasing the piece to your choir — even small rhythm errors only show up when you start singing along.
• For important pieces ask the publisher for the original MusicXML — perfect quality saves correction work later.