The R&D team at Spotify has released Basic Pitch, a free, open source tool for converting audio files to MIDI format. Basic Pitch, which uses machine learning to transcribe musical notes, can convert a recording of almost any instrument — including voice — to a MIDI version.
Over the last 40 years, musicians who use computers to compose, produce, and perform their music have primarily used MIDI, a digital standard that acts as sheet music for computers, describing which notes play and when in an easy-to-edit format. Despite its widespread use, creating a composition from scratch with MIDI can be difficult. MIDI notes are typically generated by musicians using an interface, such as a keyboard designed for that purpose, or by manually typing the notes into their software.
Live performances with real instruments are frequently difficult for a computer to interpret, which is a problem for singers who are unfamiliar with piano keyboards or complex software. Even for musicians who are familiar with MIDI composition, it can be a time-consuming process.
“To solve this problem,” Spotify says. “Researchers at Spotify’s Audio Intelligence Lab teamed up with our friends at Soundtrap to build Basic Pitch — a machine learning model that turns a variety of instrumental performances into MIDI.”
Basic Pitch has advantages over other note-detection systems, such as pitch bend detection and tracking multiple notes across multiple instruments at the same time. Basic Pitch is also low on resources, allowing it to run faster than other systems of its type on most modern computers.
Basic Pitch aims to provide musicians and producers with “power and flexibility” of MIDI without the need for specialised equipment, allowing them to record whenever inspiration strikes and edit their compositions later.