• SoX – Sound eXchange, the Swiss Army knife of audio manipulation
  • FFmpeg – ffmpeg video converter
    • -ss: the time offset from beginning. (
    • -t duration: record or transcode duration seconds of audio/video.
    • -vn no video: ignore video stream
  • FFprobe – complement to FFmpeg, gives information on files.
  • Audacity – UI software to inspect audio file, convenient to see their audio waves.


See also phab:T251638.

Extract noise samples

Given a file with silence in its first 0.3sec. :

# sox in.ext out.ext trim {start:} {duration:}
sox audio.wav noise-audio.wav trim 0 0.300     
# or :
ffmpeg -i audio.wav -vn -ss 00:00:00 -t 00:00:00.300 noise-audio.wav

For a folder with .wav audios with noisy silence in first 150ms and last 200ms :

# Create noise samples
mkdir -p ./noise;
for file in ./wav/*.wav;
do key=$(basename "$file" .wav);
  ffmpeg -i "$file" -vn -ss 00:00:00 -t 00:00:00.150 ./noise/noise-"$key"-head.wav ;
  ffmpeg -vn -sseof -0.200 -t 0.200 -i "$file" ./noise/noise-"$key"-tail.wav
  # sox "$file" ./noise/noise-"$key"-head.wav trim 0 0.200     
  # sox "$file"  ./noise/noise-"$key"-tail.wav reverse trim 0 0.200 reverse

Generate a noise profile in sox

Background audio noise sample from Shtooka project based on 282 samples from 191 audio files' header (0.17s) and tail (0.23s), totalling 51 sec. of noise. Such file are used as sample to remove background noise (denoise) in audio files recorded in similar conditions.
# From one noise sample audio file
sox noise-audio.wav -n noiseprof
# Contact noise samples into single noise sample
ffmpeg -f concat -safe 0 -i <(for f in ./noise/*.wav; do echo "file '$PWD/$f'"; done) -c copy ./noise/noise-audio.wav

# From sample, creates noise profile
sox ./noise/noise-audio.wav -n noiseprof ./noise/ ;

Clean the noise from the audio

Single file :

sox audio.wav audio-clean.wav noisered 0.21

Batch of files:

# Denoise
mkdir -p ./clean;
for file in ./wav/*.wav;do key=$(basename "$file" .wav);
sox "$file" ./clean/"$key".wav noisered ./noise/ 0.21;

According to source :

Change 0.21 to adjust the level of sensitivity in the sampling rates (I found 0.2-0.3 often provides best result).



See also phab:T251640.

To fade-in on 0.3s and out in 0.4s, you can use a simple ./ bash script, such as:

sox input.wav output.wav fade "0:0.3" `soxi -d input.wav` "0:0.3"

Or as a script:

#! /bin/bash

LENGTH=`soxi -d $WAV_IN`


soxi -d returns the length of the wav file. See sox documentation for more on soxi.

You can run this bash script as follows:

./ input.wav faded.wav



Content to create. See also phab:T213535 and LinguaLibre:Chat_room#Normalize_loudness.

See also

Lingua Libre technical helps
Template {{Speakers category}} • {{Recommended lists}} • {{To iso 639-2}} • {{To iso 639-3}} • {{Userbox-records}} • {{Bot steps}}
Audio files How to create a frequency list?Convert files formatsDenoise files with SoXRename and mass rename
Bots Help:BotsLinguaLibre:BotHelp:Log in to Lingua Libre with PywikibotLingua Libre Bot (gh) • OlafbotPamputtBotDragons Bot (gh)
MediaWiki MediaWiki: Help:Documentation opérationelle MediawikiHelp:Database structureHelp:CSSHelp:RenameHelp:OAuthLinguaLibre:User rights (rate limit) • Module:Lingua Libre record & {{Lingua Libre record}}JS scripts: MediaWiki:Common.jsLastAudios.jsSoundLibrary.jsItemsSugar.jsLexemeQueriesGenerator.js (pad) • Sparql2data.js (pad) • LanguagesGallery.js (pad) • Gadgets: Gadget-LinguaImporter.jsGadget-Demo.jsGadget-RecentNonAudio.js
Queries Help:APIsHelp:SPARQLSPARQL (intermediate) (stub) • SPARQL for maintenance (stub) • Lingualibre:Wikidata (stub) • Help:SPARQL (HAL)
Reuses Help:Download datasetsHelp:Embed audio in HTML
Unstable & tests Help:SPARQL/test
Categories Category:Technical reports