The MossFormer2_SS_16K model weights for 16 kHz speech separation in ClearerVoice-Studio repo.
This model is trained on large scale datasets inclduing open-sourced and private data.
It separates mixed-speaker speeches into individual speaker's speech.