The dataset consists of 56.7 hours of nonverbal vocal sound clips from 1419 speakers in South Korea, with metadata such as age, sex, authenticity level, and noise labeled to each utterance. It includes 16 classes of nonverbal sounds like coughing, laughing, and screaming.