Free Music Archive Dataset

dataset·creative commons music audio·active

Full-length Creative Commons music with audio, features, metadata and hierarchical genre labels for MIR.

Recorded facts

Official site	https://github.com/mdeff/fma ↗
Geography	Global
arxiv	1612.01840
github repository	mdeff/fma
size	albums: 14854 · tracks: 106574 · artists: 16341 · audio gib: 917 · duration days: 343
scope	Creative Commons music for genre recognition and other MIR tasks.
genres	161-genre hierarchy
license	Per-track Creative Commons licenses; code and metadata terms are separate.
steward	FMA dataset authors
creators	Michaël Defferrard; Kirell Benzi; Pierre Vandergheynst; Xavier Bresson
languages	multilingual metadata
modalities	MP3; CSV metadata; precomputed features; tags; free text
geographies	Global
access method	GitHub code and documented dataset archives.
intended uses	genre classification; MIR benchmarking; representation learning
related papers	FMA: A Dataset for Music Analysis
annotation method	FMA metadata and tags plus precomputed features; authors supply benchmark splits.
collection method	Curated from the Free Music Archive catalogue.
provenance claims	Paper describes the audio as Creative Commons licensed.
current availability	Repository active; mirrors and original audio availability can vary.
commercial use limits	Varies by track; some Creative Commons licenses prohibit commercial reuse.
us market scope basis	Publicly available or materially used in US-facing music/audio-AI research.
limitations biases disputes	Catalogue and genre distribution reflect the historical Free Music Archive community.
redistribution restrictions	Must comply with each track's license and attribution requirements.