nemo/collections/audio/README.md

Audio processing collection

The NeMo Audio Collection supports a range of models tailored for audio processing tasks, including single- and multi-channel speech enhancement and restoration.

• Mask-based speech processing: single-channel masking and guided source separation (GSS)
• Predictive speech processing: NCSN++
• Score-based generative models: SGMSE+
• Schrödinger bridge-based models
• Flow-matching-based models
• Multi-channel audio processing: mask-based beamforming (MVDR) and dereverberation (WPE)

More details can be found in NeMo documentation.