MedleyDB

dataset·multitrack music·active

Annotated royalty-free multitrack recordings supporting melody, pitch, instrument and source-separation research.

Verified facts

Official site	https://github.com/marl/medleydb ↗
Geography	United States
github repository	marl/medleydb
size	original release tracks: 122 · mirdata melody subset tracks: 108
scope	Multitrack songs with annotation-intensive MIR ground truth.
genres	multiple
license	Mixed; many tracks CC BY-NC-SA 4.0
steward	NYU Music and Audio Research Laboratory
creators	Rachel Bittner; Justin Salamon; Mike Tierney; Matthias Mauch; Chris Cannam; Juan Pablo Bello
versions	1.x; 2.0 additions
languages	multiple/unknown
modalities	multitrack audio; mixtures; metadata; melody f0; pitch; instrument activations
geographies	Global
access method	Metadata and annotations on GitHub; audio access requires the dataset package or request routes.
intended uses	melody extraction; instrument recognition; pitch tracking; source separation
related papers	MedleyDB: A Multitrack Dataset for Annotation-Intensive MIR Research; MedleyDB 2.0
annotation method	Expert and derived annotations for melody, pitch and instrumentation; version-controlled corrections.
collection method	Royalty-free multitrack recordings contributed or licensed from multiple sources.
provenance claims	Project describes the recordings as royalty-free; track-specific terms vary.
current availability	Annotations active; audio is available upon request in mirdata documentation.
commercial use limits	Many tracks prohibit commercial use.
us market scope basis	Publicly available or materially used in US-facing music/audio-AI research.
limitations biases disputes	Known audio and metadata issues are tracked separately.; License terms are not uniform across all tracks.