Rikorose/DeepFilterNet

Health v2 · maintenance only

Flags: no_license

Adoption not part of the score

4632 stars · 506 forks observed · 2026-08-28

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

DeepFilterNet is a low-complexity speech enhancement framework that performs real-time noise suppression on full-band 48kHz audio using deep filtering neural networks. It ships as a Python/PyTorch library, a pre-compiled Rust CLI binary, and a LADSPA plugin for use as a virtual noise-suppression microphone.

Use cases

• remove background noise from wav recordings
• real-time microphone noise suppression on linux with pipewire
• denoise audio files from the command line
• enhance speech for calls or streaming
• run speech enhancement on embedded devices
• train a custom speech enhancement model

When to choose

• you need low-latency, real-time noise suppression for full-band 48kHz audio
• you want a lightweight Rust binary without Python dependencies
• you need a system-wide virtual noise-suppression microphone via LADSPA/PipeWire
• you want to experiment with or train speech enhancement models in PyTorch

When to avoid

• your audio is not 48kHz wav (the CLI only supports that)
• you need general audio processing beyond speech enhancement
• you require a fully permissive license (the repo license is non-standard)
• you need GPU-accelerated batch processing via the CLI (GPU requires the Python backend)

Facets

library · maturity active

audio-processing machine-learning deep-learning speech-recognition machine-learning deep-learning speech-processing windows python rust noise-suppression speech-enhancement deep-filtering real-time-audio ladspa-plugin pytorch fullband-audio audio linux macos

Member repositories

For agents

markdown · JSON · MCP: product_card(name="Rikorose/DeepFilterNet")