I am looking for a way to normalize the sound in MANY MP3 files I have. Some have low sound, while others are louder so I have to turn the volume up or down depending on the song. What ways are there to do this for all files. I specially would like to do it via the terminal but GUI ways are also accepted.

  • For normalizing while playing see: askubuntu.com/questions/95716/…. This will however not change your mp3 file content - may be of advantage ;) – Takkat Jan 22 '13 at 18:09
  • Not while playing, don't want to have to set it everytime or have a player set the normalize every time I want to listen to the songs. For example, let us say I want to copy the songs to an ipod shuffle or to a pen drive to listen in a mp3 capable player. – Luis Alvarado Jan 22 '13 at 18:13
  • @Takkat BTW nicely done in that other question. Good information. – Luis Alvarado Jan 22 '13 at 18:24
  • 1
    My wife just tested the Audacity method, and it worked perfectly! Recommended. Please, when people ask for advice it is because they are NOT experts. So don´t tell them to use command line tools when they can do the job with easy-to-understand graphical tools like Audacity. Telling new Linux users to open terminal and run command line tools will just scare them away from Ubuntu, with a feeling that Windows is simple, Linux is difficult. It is not a surprise that DOS is dead, but Windows is alive. – user297240 Jun 23 '14 at 20:42
  • 1
    Scaring away users who find typing words into a "command prompt" too difficult has no downsides. – rupert Apr 15 '16 at 0:41

Audacity Install audacity

With Audacity we can easily batch process files to apply conversions or effects to many files in a list. To do so we first have to define a "Chain" containing the effects we want to apply.

This is done with "File --> Edit Chains...". In the now opening window press the Add button on the bottom left to insert a new chain (give it a sensible name):

enter image description here

Then choose the effect and it's parameters to insert to the chain (here shown for default values and the Normalize effect).

Important: we always need to also add the effect "Export MP3" (or any other export format) for saving the resulting conversion to disk.

When done leave this window with OK to open "File --> Apply Chain...". Select the Chain we have just created and load all files you need with "Apply to files...". Several files can be selected from the file chooser that opens.

enter image description here

Processed files will be saved in a new subdirectory "cleaned" in the original's path.

Sox Install sox

From version > 14.3 we can use the sox filter --norm for normalizing audio on the command line or for batch processing:

sox --norm infile outfile

MP3-support is added to Sox with libsox-fmt-all Install libsox-fmt-all.

  • 2
    Do these tools decode and re-encode? – qed Oct 29 '14 at 21:59
  • You can't normalize without re-encoding, at least the levels... – Takkat Oct 29 '14 at 22:03
  • For non-destructive live-normalizing using LADSPA and pulseaudio see askubuntu.com/questions/95716/… – Takkat Oct 29 '14 at 22:08
  • This does re-encode, so it reduces quality. See my answer if you want to preserve your file quality. Re-encoding is not necessary if your player supports volume tagging. – Wil May 16 at 6:21

Take a look @ mp3gain which for me is even better than normalize-audio

mp3gain -r *.mp3

another useful version could be the -c which prevent to ask if you want to do the changes for many files:

mp3gain -c -r *.mp3

as said in the man page:

mp3gain does not just do peak normalization, as many normalizers do. Instead, it does some statistical analysis to determine how loud the file actually sounds to the human ear. Also, the changes mp3gain makes are completely lossless. There is no quality lost in the change because the program adjusts the mp3 file directly, without decoding and re-encoding.

Note: That package was removed on purpose on ubuntu 15.04.

Debian proposes the python-rgain package as replacement (The advantage is that 'replaygain' supports several file formats, namely Ogg Vorbis , Flac, WavPack and MP3. Also, it allows you to view existing Replay Gain information in any of those file types). After installing it, run replaygain.

To install python-rgain from the terminal, run the command

sudo apt-get install python-rgain

Alternatively, get the .deb file for 14.04 (the latest) from here. Install as usual. After that, you need to run sudo apt-get -f install to correct some dependencies issues.

  • 1
    Also, if you don't want to use the terminal, there is a GUI available for it named easymp3gain-gtk, which makes it very handy! – gilbertohasnofb Nov 12 '14 at 21:26
  • This is super useful. I was very impressed by the quality of normalization using MP3Gain's Windows GUI, so I was glad to find this answer when I needed a Linux command line solution. Would definitely recommend to others. – Alex P. Miller Feb 5 '15 at 18:05
  • Could you please add some information on how to install it? It doesn't come with Ubuntu by default, and I can't find the package. – Błażej Michalik May 3 '16 at 17:36
  • Thanks... installed python-rgain without any dependency errors on Ubuntu 16.04. – Bharat Mallapur Oct 13 '18 at 6:47

I would use this project Normalize, it's a command-line tool for normalizing audio files. Looks to be exactly what you need. Can do batch processing and doesn't require resampling to intermediate formats.

It's in the package repos as normalize-audio, sudo apt-get install normalize-audio. This is a build maintained upstream by Debian so it should be in anything LTS or newer and is built with mp3 compatibility (tested). There is a good manpage man normalize-audio to explore the options but the commands defaults appear to work well. For batch processing (normalize volume across multiple files), normalize-audio -b *.mp3 or specify individual filenames instead of using wildcards.

  • The OP wants instructions for doing this. Can you provide those? – Seth Jan 22 '13 at 18:13
  • @iSeth I was initially incorrect about the source only comment because I couldn't find it with apt-cache search. I've updated with details about the deb package. – sean_m Jan 23 '13 at 8:23
  • This seems a good alternative but can't find the correct encoder and get rid of "no encoder available". Tried with libsox-fmt-mp3, libavcodec-extra. -b *.mp3 do something with only one (random?) file. – Pablo Bianchi Mar 22 '17 at 19:04


Faster and easy replaygain:

This package provides a Python package to calculate the Replay Gain values of audio files and normalize the volume of those files according to the values. Two basic scripts exploiting these capabilities are shipped as well.

Replay Gain is a proposed standard designed to solve the very problem of varying volumes across audio files.

Install: sudo apt install python-rgain.

replaygain --force *.mp3
  • -f, --force Recalculate Replay Gain even if the file already contains gain information.

Since only calculate/change replaygain value, is also faster: With an average PC (Intel i7-6500U, 8GB RAM) the rate was ~20 files/minute.



For the sake of it, I'll throw my 2 cents in. I was looking for exactly the same thing (only for ogg files) and started a thread at Crunchbang Forum. You can view it here: Normalize-audio can't find mp3 decoder

Basically my solution was the script in post #8. It works for mp3, flac, and ogg input files, possibly others but definitely not wav.

Just create a file (name it whatever you want, I called mine db_adjust_mp3), chmod +x , and stick it in your ~/bin folder. It fills in any missing codec data as well. Example:

Original file: 16._This_Protector.mp3: Audio file with ID3 version 2.3.0, contains:


Normalized file: 16._This_Protector.mp3: Audio file with ID3 version 2.3.0, contains: MPEG ADTS, layer III, v1, 192 kbps, 44.1 kHz, JntStereo

I've modified the script to use normalize-mp3 here so you can use it if you want:


find "$1" -iname "*.""$2" > $HOME/file_list

cat $HOME/file_list | while read line; do
#  echo "$line"
  orig_gain="$(normalize-mp3 -n "$line" | cut -d 'd' -f1)"
  larger=$(echo "$orig_gain"'>'-12 | bc)
  if [[ larger ]]
      gain_difference=$(echo "$orig_gain"*-1-12 | bc)
      gain_difference=$(echo "$orig_gain"-12 | bc)
  echo "Gain Difference will be: $gain_difference""db"
  normalize-ogg --mp3 --bitrate "$3" -g "$gain_difference""db" -v "$line"

This script calculates the difference between the current db level and -12db, then applies a gain adjustment to put the gain at exactly -12db, which is what I've found works the best for me. It is recursive as well, which makes it great for doing entire music collections or files in many subfolders. If you wish to set a different db level, just change the both instances of the number "12" to whatever db level you would like to use. As I posted in my Crunchbang thread, usage is as follows:

normalize-mp3 <directory> <file extenstion(with no leading dot)> <bitrate>

However, when I used to keep my music library in mp3 format, I used to use mp3gain as well, just as Philippe suggested. The dead simplicity of it is great and I really liked it. The problem with normalize-audio though is that it does decode an re-endcode the files, so there is some sound degradation. But unless you're an audiophile and your mp3's are encoded at a high bitrate you shouldn't notice much difference.

The thing I noticed with mp3gain though was that no matter what options I tried I couldn't get everything in my collection to be exactly the same db level, which is what I want so that I never have to adjust the volume from one track to the next. This script does exactly that. Sorry for being so long winded. Hope this helps.


I liked Neil's answer the most, because it doesn't introduce correlation between audio files: just pick one gain level and adjust everything to it.

However I had some problems parsing the output of normalize-ogg with some files I have. There is also one nasty issue with bc: it doesn't do real rounding, it only truncates.

So eventually I gave up on shell scripting and moved to python.

Note1: the exiftool part might be overkill but I wanted to be 100% sure that the original bitrate would be preserved.

Note2: this will overwrite the originals, if you want to preserve them, use --backup in the last call to normalize-ogg. But I found more practical to keep a copy in a separate, safer, directory.

Note3: this solution deals with ogg files, but it's trivial to adapt it to mp3, just replace the occurrences of "ogg" with "mp3".

Here's my take at the problem. The latest version can be found here: regain.py

Parallel normalize gains
    This program is free software: you can redistribute it and/or modify
    it under the terms of the GNU General Public License as published by
    the Free Software Foundation, either version 3 of the License, or
    (at your option) any later version.

    This program is distributed in the hope that it will be useful,
    but WITHOUT ANY WARRANTY; without even the implied warranty of
    GNU General Public License for more details.

    You should have received a copy of the GNU General Public License
    along with this program.  If not, see <http://www.gnu.org/licenses/>.

# Absolute value, in dB for the desired gain of each file


from subprocess import Popen, PIPE
from multiprocessing.dummy import Pool as ThreadPool
from os import listdir
import logging

def initlogger(logfile="log.log", mainlevel=logging.DEBUG,
               filelevel=logging.DEBUG, consolelevel=logging.DEBUG):
    # create logger 
    logger = logging.getLogger()
    # create file handler which logs even debug messages
    fh = logging.FileHandler(logfile)
    # create console handler also logging at DEBUG level
    ch = logging.StreamHandler()
    # create formatter and add it to the handlers
    formatter = logging.Formatter("%(asctime)s [%(threadName)-12.12s] [%(levelname)-5.5s]  %(message)s")
    # add the handlers to the logger

def logcommand(command=[]):
    if not isinstance(command, list):
        return "", "", -1
    logging.info("Command:\n" + " ".join(command) + "\n")
    proc = Popen(command, stdin=PIPE, stdout=PIPE, stderr=PIPE)
    output, err = proc.communicate()
    output = output.decode("utf-8")
    err = err.decode("utf-8")
    logging.info("Output:\n" + output + "\n")
    logging.info("Error:\n" + err + "\n")
    logging.info("Return Code:\n" + str(proc.returncode) + "\n")
    return output, err, proc.returncode

def regain(target):
    logging.info("============================ Start File  ============================")
    logging.info("Extracting gain info.\n")
    commandgetlevels = ['normalize-ogg', '-n', target["name"]]
    output, err, retcode = logcommand(commandgetlevels)

    level  = output.split()[0]
    logging.debug("Level: " + level)
    if "dBFS" in level:
        level = level.split("dBFS")[0]
    level = level.replace(',', '.')
    level = int(round(float(level)))
    delta = target["gain"] - level
    logging.info("Required adjustment: " + str(delta) + "\n")
    if delta is 0:
        logging.warning(target["name"] + " is already at the correct level")
        return 0

    logging.info("Extracting average bitrate.\n")
    commandgetinfo = ['exiftool', target["name"]]
    output, err, retcode = logcommand(commandgetinfo)
    bitrate = '0'
    for line in output.split('\n'):
        if 'Nominal Bitrate' in line:
            bitrate = line.split(':')[1].split()[0]
    logging.info("Average bitrate is: " + str(bitrate) + "\n")
    if bitrate is '0':
        logging.error("No valid bitrate found, aborting conversion.\n")

    commandrenormalize = ['normalize-ogg', '--ogg', '--bitrate', bitrate,
                          '-g', str(delta) + 'db', target["name"]]
    output, err, retcode = logcommand(commandrenormalize)
    if retcode is not 0:
        log.error("Output:\n" + output)
        log.error("err:\n" + err)

    return retcode

# function to be mapped over
def parallelregain(gain=TARGET_GAIN, threads=MAX_THREADS):
    logging.info("Creating thread pool with " + str(threads) + " elements.\n")
    pool = ThreadPool(threads)
    targets = []
    files_list = listdir(".")
    counter = 0
    for filename in files_list:
        if filename.endswith("ogg"):
            target = {
            counter = counter + 1
    pool.map(regain, targets)

if __name__ == "__main__":
    initlogger(logfile="normalize.log", consolelevel=logging.WARNING)

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.