MagnaTagATune

dataset·music autotagging·active

Magnatune audio clips with human tags collected through the TagATune game for music auto-tagging research.

Recorded facts

Official site	https://mirg.city.ac.uk/datasets/magnatagatune/index1.html ↗
Geography	Global
size	clips: 25863 · reported tags: 188 · clip duration seconds approx: 29
scope	Short music clips and crowd-generated semantic tags.
genres	Magnatune catalogue
license	Noncommercial redistribution/repurposing claim; exact corpus license not normalized.
steward	City University Music Informatics Research Group mirror
creators	Edith Law; Olivier Gillet; Kris West; Michael Mandel; Mert Bay; J. Stephen Downie
languages	primarily English tags
modalities	16 kHz mono MP3; CSV; database dump; human tags; audio features
geographies	Global
access method	City University mirror after the original TagATune host went offline.
intended uses	music auto-tagging; multi-label classification
related papers	Evaluation of algorithms using games: the case of music annotation
annotation method	Human computation via annotation and comparison rounds.
collection method	Audio excerpts from Magnatune; annotations collected through an online agreement game.
provenance claims	TagATune states source artists used similar licenses allowing noncommercial redistribution and repurposing.
current availability	Mirror page available.
commercial use limits	Commercial reuse is not established by the mirror.
us market scope basis	Publicly available or materially used in US-facing music/audio-AI research.
limitations biases disputes	Original host went offline.; Tags reflect game participants and the Magnatune catalogue.
redistribution restrictions	Underlying clip terms and attribution links must be preserved.