This dataset contains preprocessed 4-second audio chunks from the Choral Singing Dataset (CSD) formatted for SATB (Soprano, Alto, Tenor, Bass) voice source separation tasks.
This is the compressed version with 8kHz sample rate and int8 precision for smaller file sizes.