Overview

NVIDIA Background Noise Removal (BNR) NIM is an audio background noise removal model from NVIDIA. BNR NIM removes unwanted noises from audio, improving speech intelligibility and also improving the speech recognition accuracy of various ASR systems in noisy environments.

NVIDIA Background Noise Removal NIM models are built on the NVIDIA software platform, incorporating CUDA, TensorRT, and Triton to offer out-of-the-box GPU acceleration.

NVIDIA Background Noise Removal (BNR) NIM has two modes:

• Streaming mode uses real-time peer-to-peer audio processing with a continuous data flow, making it ideal for live applications like video conferencing and live broadcasting.

  In this mode, raw audio data is streamed to the NIM in chunks of 10 ms, and the NIM streams back the processed data in real time.

• Transactional mode processes complete audio files in a single request-response interaction, making it better suited for offline audio enhancement and post-production workflows.

Note

The difference in modes is limited to the size of audio inputs and outputs and the cadence at which audio is processed by the NIM. Quality of outputs is the same for both modes.

Architecture

NVIDIA Background Noise Removal uses a residual convolutional recurrent neural network (CRNN) with sequential modeling applied to the encoded latent representation.

Try It Out

Try out the NVIDIA Background Noise Removal NIM at nvidia/bnr.

Additionally, access the Try API feature to experience the NVIDIA Background Noise Removal NIM API without hosting your own servers, because Try API uses the NVIDIA Cloud Functions backend.