vidtoolz-rnnn

Noise reduction using ffmpeg rnnn model

Installation

First install vidtoolz.

pip install vidtoolz

Then install this plugin in the same environment as your vidtoolz application.

vidtoolz install vidtoolz-rnnn

Usage

type vid rnnn --help to get help

Basic Usage

# Basic usage with explicit output
vidtoolz rnnn -i input.mp3 -o output.wav -m bd --mix 0.6

# Automatic output filename generation
vidtoolz rnnn -i input.mp3 -m lq

# With custom mix ratio
vidtoolz rnnn -i input.mp3 -o output.wav -m mp --mix 0.8

Command Line Arguments

• -i INPUT_AUDIO, --input_audio INPUT_AUDIO Path to input audio file (e.g. mp3, wav) - Required

• -o OUTPUT_WAV, --output_wav OUTPUT_WAV Path to output WAV file - Optional (auto-generated if not provided)

• -m {bd,cb,lq,mp,sh}, --model {bd,cb,lq,mp,sh} RNNoise model to use. Available models: bd, cb, lq, mp, sh - Required

• --mix MIX Wet/dry mix ratio (0.0 = original, 1.0 = fully denoised) - Optional (default: 0.6)

Available Models

The plugin automatically discovers available RNNoise models from the models directory:

• bd - beguiling-drafter-2018-08-30
• cb - conjoined-burgers-2018-08-28
• lq - leavened-quisling-2018-08-31
• mp - marathon-prescription-2018-08-29
• sh - somnolent-hogwash-2018-09-01

Examples

# Denoise a podcast recording with studio model
vidtoolz rnnn -i podcast.mp3 -m sh -o podcast_clean.wav

# Clean up a noisy street interview
vidtoolz rnnn -i interview.mp3 -m mp --mix 0.7

# Process multiple files (bash loop)
for file in *.mp3; do
    vidtoolz rnnn -i "$file" -m cb
 done

RNNoise Model Suitability Summary

Development

To set up this plugin locally, first checkout the code. Then create a new virtual environment:

cd vidtoolz-rnnn
python -m venv venv
source venv/bin/activate

Now install the dependencies and test dependencies:

pip install -e '.[test]'

To run the tests:

python -m pytest