SVDD Challenge 2024

dataset·singing voice deepfake detection·closed

The inaugural controlled and in-the-wild singing-voice deepfake detection challenge held with IEEE SLT 2024.

Recorded facts

Official site	https://main.singfake.org/ ↗
Geography	Global
arxiv results	2408.16132
arxiv evaluation plan	2405.05244
size	participating teams controlled track: 47
scope	Detection of bona fide versus synthetic singing in controlled and web-collected conditions.
genres	multiple
license	research/challenge terms
steward	SVDD Challenge organizers / University of Rochester-led collaboration
creators	You Zhang; Yongyi Zang; Jiatong Shi; Ryuichi Yamamoto; Zhiyao Duan
languages	multiple singing languages
modalities	singing audio; binary labels; attack metadata; leaderboard scores
geographies	Global
access method	Challenge site; research access to associated datasets.
intended uses	singing deepfake detection evaluation
consent claims	Source-singer and web-platform consent is not uniformly documented.
related papers	SVDD 2024: The Inaugural Singing Voice Deepfake Detection Challenge
prohibited uses	Identity abuse, impersonation or use outside dataset terms.
annotation method	Binary provenance labels and attack partitions support equal-error-rate evaluation.
collection method	Combines CtrSVDD generated examples with WildSVDD in-the-wild examples.
provenance claims	Official site and evaluation paper describe both tracks and datasets.
current availability	Challenge closed; documentation remains online.
commercial use limits	Commercial use not established.
us market scope basis	Publicly available or materially used in US-facing music/audio-AI research.