Basic Pitch’s machine learning model at IEEE ICASSP 2022

Опубликовано: 09 Июль 2026
на канале: Spotify R&D
2,518
88

Do ML models have to be big to be useful? Rachel Bittner, Senior Research Scientist in Spotify’s Audio Intelligence Lab, presents the research behind Basic Pitch, a free open source tool for converting audio into MIDI.

Basic Pitch uses machine learning to transcribe the musical notes in a recording. Unlike similar ML models, Basic Pitch is not only versatile and accurate, but also fast and computationally lightweight. It works with recordings from almost any single instrument, including voice, and also supports pitch bend detection. It was built for artists and producers who want an easy way to turn their recorded ideas into MIDI, a standard for representing notes used in digital music production.

Try the Basic Pitch demo at: https://basicpitch.io

Learn more on the Spotify Engineering blog: https://engineering.atspotify.com/per...

Have an idea for Basic Pitch? Contribute on GitHub at: https://github.com/spotify/basic-pitch

Read the research paper about the model: https://2022.ieeeicassp.org/view_pape...

“A Lightweight Instrument-Agnostic Model for Polyphonic Note Transcription and Multipitch Estimation”

2022 IEEE International Conference on Acoustics, Speech and Signal Processing

Rachel Bittner, Juan José Bosch, David Rubinstein, Sebastian Ewert, Spotify, France; Gabriel Meseguer Brocal, Institut de recherche et coordination acoustique/musique (IRCAM), France