Description

length: 1h 32m (talking and singing) Description: the source is a night karaoke stream with ASMR-like lullaby singing style, so low-pitched vocals will be somewhat airy as supposed to be. Trained using the latest original RVC with some code tweaks, including removal of minibatch k-means code in the index training function, so the index file for 1h+ dataset won't shrink.