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.