Briefs · Taipei time (GMT+8)

NVIDIA introduces open speech-cleanup model NemotronLabs RE-USE

一句重點：NemotronLabs RE-USE is a downloadable open model with 9.6M parameters for reducing noise and echo in speech recordings.

發生什麼事

The official @NVIDIAAI account announced NemotronLabs RE-USE, a multilingual open model developed by NVIDIA researchers. It can reduce background noise, echoes and other distortions in an audio recording, helping speech sound clearer. The model has 9.6M parameters. Hugging Face has a demo, and the model is available for download. These are NVIDIA's statements (🗣 self-reported); we have not tested the model ourselves.

01Sources

  • @NVIDIAAI: NVIDIA's official AI account announced NemotronLabs RE-USE on 2026-10-30 at 20:15 UTC (04:15 Taipei time on 2026-10-31), including its features, parameter count, and Hugging Face demo and download.

More briefs →