Silence Removal Python Program

A Python utility that processes audio files using two methods: Silero VAD (Voice Activity Detection) and FFmpeg. Both approaches remove silence and normalize audio output.

Prerequisites (Other versions may be compatible, these are just what I used)

• Python 3.12.2
• FFmpeg version 2024-04-10-git-0e4dfa4709-full_build-www.gyan.dev
• Required Python packages (see requirements.txt):
  • torch==2.5.1
  • torchaudio==2.5.1

Installation

1. Install Python 3.12.2
2. Install FFmpeg and ensure it's in your system PATH
3. Install required packages:

pip install -r requirements.txt

Usage

Run the script with an input audio file (tested to work for mp3 and wav, other file formats may not be recognized):

python main.py path/to/audio/file.wav

The script will:

1. Create a 'processed' subdirectory where the input file is located. I've included a media folder with an example test file.
2. Process the file using both methods:
   • Silero VAD: {filename}_SileroVAD.wav
   • FFmpeg: {filename}_FFmpeg.wav
3. Display processing statistics

Processing Methods

Silero VAD

• ML model based voice detection
• Adds 0.5s padding around speech segments
• Minimum speech duration: 0.5s
• Handles silence by trimming to 1 second
• Uses PyTorch backend

FFmpeg

• Signal-based silence detection
• Silence threshold: -40dB
• Minimum silence duration: 1.0s
• Two-pass normalization
• Generates 1-second silent file for silent inputs

Output Statistics

For each method, displays:

• Original duration (seconds)
• Processed duration (seconds)
• Reduction percentage
• Processing time
• Number of detected segments

Error Handling

The script handles:

• Missing input files
• Processing errors
• Silent audio files
• Windows/Unix path differences
• Temporary file cleanup