Opencpop

dataset·mandarin singing voice corpus·active

Studio-recorded Mandarin singing corpus with 100 songs and phoneme, note and pitch annotations.

Verified facts

Official site	https://wenet-e2e.github.io/opencpop/ ↗
Geography	Xi'an, Mainland China, China
arxiv	2201.07429
github repository	wenet-e2e/opencpop
access note	Download requires a request form; current license terms need field-level audit.
dedupe risk	low
discovery route	Simplified Chinese, Traditional Chinese and English source discovery
documented size	hours: 5.2 · songs: 100 · utterances: 3756
evidence strength	strong
recommended prompt	06-datasets-and-benchmarks
classification rationale	Primary evidence shows material mandarin singing voice corpus relevance to AI music or audio AI.