semamba

Model type	semamba
Class	SEMambaSpeechEnhancementModel
Task	speech-enhancement
Source	models/reuse.py

Description

NVIDIA RE-USE (SEMamba) universal speech enhancement generator.

Consumes the magnitude and phase of a noisy STFT and predicts the enhanced magnitude, phase, and complex spectrogram. The magnitude is expected to already be compressed by the STFT front-end (RE-USE uses log1p); decompression and the ISTFT happen after this graph.

Usage

mobius build --model <MODEL_ID> output_dir/

from mobius import build

model = build("<MODEL_ID>")