AudioCaps

dataset·audio captions·active

Crowdsourced natural-language captions paired with AudioSet-derived audio clips.

Recorded facts

Official site	https://audiocaps.github.io/ ↗
Geography	South Korea
size	audio clips approx: 46000
scope	General audio captioning including music and musical-event clips.
genres	mixed audio; includes music
steward	AudioCaps authors
creators	Chris Dongjoo Kim; Byeongchang Kim; Hyunmin Lee; Gunhee Kim
languages	English
modalities	AudioSet clip references; human-written English captions; split metadata
geographies	Global
access method	Official data and code links.
intended uses	audio captioning; audio-language representation; music captioning transfer
related papers	AudioCaps: Generating Captions for Audios in the Wild
annotation method	Captions collected by crowdsourcing and checked for faithfulness.
collection method	Audio clips selected from AudioSet.
provenance claims	Derived from AudioSet, so audio access follows YouTube references.
current availability	Annotations available; source audio depends on AudioSet/YouTube.
us market scope basis	Publicly available or materially used in US-facing music/audio-AI research.
limitations biases disputes	General-audio rather than music-specific.; Audio availability inherits AudioSet URL decay.
related models tools benchmarks	AudioSet; WavCaps