Skip to content
alan.salmon
All projects
Completed · 2022

Silence Remover

Python app that automatically removes the silences from a WAV recording, while keeping the natural pauses of speech.

Year
2022
Context
Algorithms and data processing
  • Python
  • Signal processing
  • Matplotlib
Project poster: a sound wave cut in the middle, on a waveform pattern

01 · Context

The challenge

Personal signal-processing project, with no external audio library: the WAV file is read byte by byte and converted by hand into 16-bit samples, analysed, then rebuilt without its silences. A small Kivy window lets you drop in the file to process.

02 · Architecture

How it's built

Processing diagram: the WAV file dropped into the Kivy window is read as 16-bit samples, silences longer than 400 ms are detected under a threshold and cut with a margin kept, and the cleaned file is written back
The processing chain, and a closer look at silence detection

Key points

  • WAV reading and writing done by hand: bytes converted into signed 16-bit samples, and back
  • Silences detected with a threshold on the normalised signal, only beyond 400 ms so that the natural pauses of speech are kept
  • Short unwanted noises (6 ms at most) merged into the surrounding silence, and a 100 ms margin kept around each cut
  • Thresholds tuned by plotting the signal and the detected sections with Matplotlib

03 · In pictures

Gallery

  • Illustration: a waveform with three silent sections under a dotted threshold highlighted, then the same wave stitched back together without them
    The idea: silences found under the threshold are removed and the sound is stitched back together
  • Application window asking to drop a WAV file
    The application window, waiting for a file
  • Application window showing “Fichier généré” (file generated) and the name of the output file, test1_OUT.wav
    The cleaned file is written next to the original
  • Matplotlib chart of the samples normalised between −1000 and 1000, with the two horizontal thresholds and red and yellow vertical lines marking the start and end of each removed silence
    Tuning with Matplotlib: silence thresholds and removed sections

Click an image to open it full screen and browse through the screenshots.

04 · Takeaways

What I took away from it

This personal project introduced me to Kivy and sparked my interest in scripting. Since then, I’ve enjoyed writing little Python scripts to automate small, repetitive everyday tasks on a computer.