xiph/rnnoise

Health v2 · maintenance only

Adoption not part of the score

5801 stars · 1071 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

RNNoise is a C library that uses a hybrid DSP/recurrent neural network approach for real-time full-band speech noise suppression. It also ships with a simple command-line example tool that denoises raw 16-bit mono PCM files at 48 kHz.

Use cases

• remove background noise from microphone audio in real time
• denoise raw PCM speech recordings
• integrate noise suppression into a VoIP or streaming application
• clean up noisy speech before speech recognition
• train a custom noise reduction model with my own speech and noise data

When to choose

• you need lightweight, real-time noise suppression in C with low CPU overhead
• you want a proven BSD-licensed denoiser you can embed in audio pipelines
• you work with 48 kHz speech audio and want full-band enhancement

When to avoid

• you need WAV or compressed audio handling out of the box (the CLI only takes raw PCM)
• you need music or general audio denoising rather than speech enhancement
• you want a high-level Python or GPU-accelerated deep learning denoiser

Facets

library · maturity stable

audio-processing machine-learning speech-recognition machine-learning speech-processing windows cpp cross-platform noise-suppression rnn dsp denoising speech-enhancement real-time-audio audio linux macos

Member repositories

For agents

markdown · JSON · MCP: product_card(name="xiph/rnnoise")