Radar ·
NVIDIA RE-USE speech enhancement model arrives on MLX-Audio
A post dated November 2, 2026 reports that RE-USE, a small multilingual universal speech enhancement model from NVIDIA, is now available on MLX-Audio to clean up noisy recordings.
Radar records that NVIDIA's RE-USE was described as a small multilingual universal speech enhancement model, now available on MLX-Audio. According to the author of the post, dated November 2, 2026, the model helps clean noisy recordings that speech-to-text models often struggle with. The post also suggests that a lightweight MLX speech enhancement model could improve ASR input for noisy audio on local hardware.
The source is a third-party post and gives no benchmarks, quality metrics, hardware requirements or usage instructions. To confirm these points, consult the original post and the MLX-Audio documentation through the link kept in the Radar archive, and check the license before any use. The relevance to Rota Nacional is indirect: the platform offers audio transcription that already passes through the same personal data detection and organizational policy barrier, but the speech enhancement described here is not a Rota feature.
Original source: Modelo de melhoria de fala RE-USE da NVIDIA agora disponível no MLX-Audio