From “zero to slice”: Soundslice takes on optical music recognition with AI

Soundslice syncs audio, video and music notation into a single web-based interface that can be used on virtually any piece of music. A favorite tool for musicians that want to learn how to play music by interacting with both the sheet music and a recording, it’s at the cutting edge of some developments in the more general field of music applications.
As it has rolled out more robust features over time, Soundslice’s goal has been to get from “zero to slice” as quickly as possible. A “slice”, in Soundslice parlance, is any piece of music notation, large or small.
One crucial way of getting to that slice is from existing music that may not exist as a recording or a music notation format that can be easily converted to MusicXML. So that’s where music scanning comes into play.
Although music scanning, or optical music recognition (OMR) technology has been around for decades, it’s now possible to marry music-notation know-how with advancements in machine learning, or “artificial intelligence”. Soundslice’s latest feature aims to harness that technology to learn as it goes, and get better at interpreting the music with every scan it processes.
Soundslice’s founder, Adrian Holovaty, returns to Scoring Notes to talk all about how Soundslice works in this regard and helps us understand its potential to revolutionize this important step in music preparation.
More reading and listening:

Adrian Holovaty: Thoughts on my first machine learning project
Soundslice: State-of-the-art music scanning
Scoring Notes: Soundslice’s Adrian Holovaty
Scoring Notes: Marie Chupeau and the human side of Newzik’s artificial intelligence