Million Song Dataset

dataset·music features metadata·active

Audio features and metadata for one million contemporary popular-music tracks, without distributed audio.

Verified facts

Official site	https://millionsongdataset.com/ ↗
Geography	United States
size	tracks: 1000000 · subset tracks: 10000 · full download gb: 280
scope	Commercial-scale popular-music metadata and derived audio analysis features.
genres	contemporary popular music
license	Mixed by field and complementary dataset
steward	LabROSA and The Echo Nest
creators	Thierry Bertin-Mahieux; Daniel P. W. Ellis; Brian Whitman; Paul Lamere
languages	multilingual metadata
modalities	HDF5 metadata; derived audio features; complementary linked datasets
geographies	Global
access method	Full download pathways and 10,000-song subset documented on official site.
intended uses	scalable MIR; recommendation; music similarity; benchmarking
related papers	The Million Song Dataset
annotation method	Automated audio analysis plus linked third-party metadata.
collection method	The Echo Nest supplied analysis and metadata; community datasets add tags, lyrics and interactions.
provenance claims	Core features and metadata provided by The Echo Nest; audio is not included.
current availability	Official site remains online; individual historical mirrors may vary.
commercial use limits	Must be determined per component and source.
us market scope basis	Publicly available or materially used in US-facing music/audio-AI research.
limitations biases disputes	No full audio is distributed.; Complementary datasets carry heterogeneous licenses and coverage.
redistribution restrictions	Audio is excluded; linked datasets have separate terms.