NVIDIA NemotronLabs RE-USE
What it is
A 9.61M parameter MambaSSM-based model for multilingual universal speech enhancement. It restores speech quality by removing noise, reverberation, and artifacts while preserving fidelity across multiple sampling rates and languages.
The problem it solves
Real-world speech recordings are often degraded by noise, echo, and compression artifacts, making them difficult to understand or process; existing enhancement models are often large, single-language, or limited to specific sampling rates.
What you could use it for
- Cleaning up noisy audio recordings from various microphones
- Removing background noise and reverberation from voice calls
- Enhancing speech clarity for ASR preprocessing
- Restoring audio quality from low-bandwidth or compressed sources
Why it matters
Significantly improves the intelligibility and quality of degraded speech recordings with a lightweight model, enabling high-quality audio processing on edge devices or consumer hardware.
Read the source ↗Back to the library →
Summary written by an AI model from the source; check the source before you rely on it.