How To Normalize Audio – Why Do It? Everything You Need To Know 100

To normalize audio is to change its overall volume by a fixed amount to reach a target level. It is different from compression that changes volume over time in varying amounts. It does not affect dynamics like compression, and ideally does not change the sound in any way other than purely changing its volume.

Why would we want to do this, what is the best way of doing it and what are the hidden dangers in terms of reducing sound quality? Let’s find out!

Why Normalize Audio?

There are only 2 good reasons to normalize:

  • GETTING THE MAXIMUM VOLUME – If you have a quiet audio file you may want to make it as loud as possible (0 dBFS ) without changing its dynamic range. This process is illustrated below (click to zoom in):
  • MATCHING VOLUMES – If you have a group of audio files at different volumes you may want to make them all as close as possible to the same volume. It may be individual snare hits or even full mixes.

Normalization can be done automatically without changing the sound as compression does. While this is a huge advantage, it can’t replace compression as it can’t affect the peaks in relation to the bulk of the sound. This means you have far less control. Often Normalizing audio just won’t work for matching volume levels, mastering engineers need not loose any sleep.

What Is The Best Method To Normalize Audio?

There are different ways of measuring the volume of audio. We must first decide how we are going to measure the volume in the first place before we can calculate how to alter it, the results will be very different depending on what method we use.

PEAK Volume Detection

This only considers how loud the peaks of the waveform are (shown in red on the right) for deciding the overall volume of the file. This is the best method if you want to make the audio as loud as possible. In digital audio you can’t get any louder than the highest peak at 0 dBFS, so normalizing to this value will create the loudest file you can.

RMS Volume Detection

This considers the ‘overall’ loudness of a file. There may be large peaks, but also softer sections. It takes an average and calls that the volume.

This method is closer to how the human ear works and will create more natural results across varying audio files.

We are still limited to the fact that digital audio can’t go above 0 dBFS. This means that to make a group of audio files the SAME volume we may need to turn them all down so that none of their peaks clip (goes over 0 dBFS). This may not be desirable, an example would be in mastering.

Another problem is that RMS volume detection is not really like human hearing. Humans perceive different frequencies at different volumes. This is shown on the Fletcher–Munson curve below:

If one sound file has many frequencies between 1000 – 6000 Hz as shown in the diagram, it will sound louder.

RMS doesn’t take this into account. Luckily there is a recent solution, the new standard in broadcast audio, the catchily titled EBU R-128


EBU R-128 Volume Detection

This is a similar way to measure volume as RMS, but can be thought EBU_R128of as emulating a human ear. It listens to the volume intelligently and thinks how we will hear it. It understands that we hear frequencies between 1000 – 6000 Hz as louder and takes that into account.

We still have the same 0 dBFS problem mentioned for RMS, but now the different normalized audio files should sound much more consistent in volume.

What Are The Hidden Dangers?

Normalization can be performed in a stand alone program, usually an audio editor (like Sound Forge), or also inside your DAW. For the sake of this section I am assuming you are using an audio editor. Inside a multi-track DAW project, when you are not exporting the normalized files individually, you probably won’t suffer from the problems I now mention.

  • Peak normalization to 0 dBFS is a very bad idea for any parts to be used in a multi-track recording. It may not clip by itself, but as soon as you add any extra processing or play tracks simultaneously your DAW or Plug-ins may overload. This subject comes under ‘gain staging’, a big subject to cover in the future.
  • It is a destructive process. Performing any digital processing to a file is going to change it. Its bad reputation was mainly earned back in the days when digital files were all stored as 16 bit. If you turned the volume down you effectively reduced the bit depth. Your CD quality 16 bit file could end up 12 bit or less. Even turning it up with PEAK normalization caused damage.

Nowadays audio editing software works internally at a much higher bit depth (often 32 bit floating point). This means that calculations are done much more accurately, and therefore affect the sound quality far less. This is only the case if we keep the file at the higher resolution once it has been processed!

To take advantage of the high quality of high bit depth inside audio editing software make sure all your temporary files are stored as 32 bit floating point. Also consider saving them in this format if you are going to do further processing.

Other Important Points To Consider

  • People often peak normalize their audio just so they can see the waveforms more clearly on the screen. This is a bad idea, your software should have an option to make the waveforms bigger without resorting to permanently altering the audio file.
  • For the matching of volume levels for finished tracks, virtual normalization is possible inside many media players (inc Foobar), the most popular is called ReplayGain. The aim of Replaygain is to try to make all the different tracks of music play back at the same volume level WITHOUT changing the actual file.It works by measuring the RMS or EBU R128 volume of a file and then deciding how much it should be turned down to match other music also using the ReplayGain system. This figure is stored inside the audio file, when it is played the software can turn the volume down itself. It’s not perfect, but it’s a very interesting method to hear different songs at the same volume level and make the loudness war a total waste of time.


In summary, normalization is a very useful tool, but also one that can easily be abused and cause an unnecessary loss of sound quality. Understanding the difference between PEAK and RMS volume is vital. Use with caution.

If you found this article useful and would like to stay in touch, please click the social media buttons below and sign up for the mailing list, I promise I will NEVER spam you!

[yikes-mailchimp form=”1″]


Leave a comment

Your email address will not be published. Required fields are marked *

100 thoughts on “How To Normalize Audio – Why Do It? Everything You Need To Know

  • Raf

    Hi thanx a lot for yr advice to this topic very helpfully, but dislike to ask you what is best for mp3s…shall I normalize them as mp3s or decode them to audio and normalize them then?Thanx in advance!!!

    • David Post author

      I am assuming you are talking about mp3 files for general listening rather than music production. If you don’t have access to the lossless original I recommend you use

      You need to have a compatible playback system, but it is very common. If you scan your entire music collection it will add data losslessly to your mp3 files that make them all play back at about the same volume. You can choose if you ‘normalize’ by album or track. I hope that helped!


    I have a question.i dont know if its a bad or unprofessional step but in the mixing stages i use a peak compressor to cut down peaks and mantain transients and keep a solo instrument in a natural sounding way if it keeps going up and own.then i use an RMS Compressor to raise the relative loudness not too much of it and then put a a limiter.without ever normalizing the audio file.i also want to understand if there are any Dynamics dangers to it.since normalization doesnt change the dynamics of a song.

    • David Post author

      The only ‘dynamic dangers’ are the ones you can hear and don’t like! So long as you can hear the changes you make at every step and like them then there are no problems.

      Using peak compression, then RMS compression, then limiting sounds OK on special occasions, but it could be overkill on a solo instrument. I would use peak compression with a long attack if you want to enhance the transient and add punch (rather than maintain the transient), RMS compression to make the sound smoother, and a limiter to ruin the sound at the end 🙂

      Seriously though. Don’t ever do anything to make something sound louder, only better! Compression can make things sound punchier, smoother and glue sounds together. It can also make things sound small, weak and pathetic. Make sure you compare before and after at the exact same volume. Check this video out for info on that

  • Milos


    any good software for RMS normalization ? I have been using for long time Foobar for normalization of MP3 collection (and it uses peak normalization if I am correct) and its overall bad.
    Lot of songs which are louder then other.

    Any good software for normalization out there ?

    • David Post author

      The newer versions of Foobar use R128 (a better version of RMS that hears more like a human), not peak, it should work very well.

      Are you using a very old version of Foobar? Also, are you choosing ‘ReplayGain > Scan per-file track gain’. is more great free software for the job, but if you are doing many mp3 files then you are better off in Foobar.

  • Zorro

    Hi David,

    1) So looking over your chart and skimming over your topic … would it be safe to say to use 89-92db as a “Standard” range for Boosting Audios in a Video MP4 without making the Audio sounding “Barrelee”?
    2) BTW, I use XMedia for my Audio Normalization!?!? What do you think about this SW?

    Thanks for the Article! 🙂

  • kelvin

    Hey! Thanks for the great post. I am wonderimg how to track vocals to send to a studio to get mixed. When i track vocals the signal seems thin or weak and i can not see it. Im afraid to normalize because i dont want to ruin the vocal take. Any suggestions?

    • David Post author

      There are a million things to do to improve vocals before and after recording, normalizing won’t affect the sound, so don’t worry about it.

      You can’t ‘see’ the signal in your DAW? Do you have enough gain in the pre-amp? It should be as loud as possible WITHOUT ever clipping (going into the red). Normalize and listen, can you see the waveform now but there is a huge amount of noise in the background? That would mean your gain was far too low.

      Best to track the vocals in the studio, otherwise you need good quality equipment in a good room with good technique and experience!

      • Kelvin

        I’m using an encore dynamic mic and tascam uh 7000 which should have enough gain. You’re right it is either the room or technique; or possibly both. Perhaps a studio is my best bet for a finalized product. Thanks for the advice and responding back.

  • Kalypso

    Hi, I have to disagree about normalizing to maximize volume for numerous reasons:
    it’s the loudness-war non-sense, that every producer feels the need to obliterate their mixes by constantly finding ways to increase loudness, it’s bad for sound. What we need is to focus on the dynamics, that’s what music is about…not how “loud” can you go.

    Another very important reason not to raise volume to maximum is when you are doing your mix down you NEED to leave lots of headroom in the audio for mastering, you cannot master an audio track that eats up all the headroom and has clipping. If you want a louder recording then turn up the levels on the mixer a bit higher or add some gain to the mixer channels. The first screenshots audio levels are perfectly fine…I don’t understand why that needs to be boosted???

    If you want a little bit of loudness, the person doing your mastering will do that for you.

    Just try to avoid the need by normalizing by recording properly in the first place 🙂

    Please enjoy your day 🙂

    • David Post author

      I agree with you 100%, but i think you are mixing up compression/limiting with normalizing. My article opens:

      “… It is different from compression that changes volume over time in varying amounts. It does not affect dynamics like compression, and ideally does not change the sound in any way other than purely changing its volume.”

      I despise the loudness war, it has ruined so many albums! Best not to digitally limit anything to gain volume, it sucks the life out of music and creates a harsh lifeless sound. You can edit the peaks down instead if there are a few offending ones. Compression on the other hand can make things sound better if done well.

      The main thing I use normalizing for is to make tracks all about the same volume in a mix (-18dBFS as a starting point), or for critical a/b listening tests. I do it with the SWS loudness tool in Reaper which is non destructive and uses the r128 integrated loudness measurement.

  • Siln

    Hi David, I have a question.
    Say one of my audio file originally exceeds the limit (0dBFS). Now that I want to remove the clipping risk, so I use Peak Normailzer to cut the peak down to somewhere near -0.3dBFS (recommended headroom level?). Is it safe? (I checked the DR Meter – unchanged, but then I realized that it only altered one side – the other side was intact. It may cause some deviation, right?)

    • David Post author

      If it went over 0dBFS then I am afraid it is already clipped. It may not be noticeable if you are lucky, so have a close listen. To be sure not to clip it again keep it under 0dBFS at all times.

      When you say -0.3dBFS I think you are talking about avoiding inter-sample peaks. This is too big a subject for my reply, but I will tell you what I do (that some might argue with). If you normalize, do it to -1dBFS. You then avoid inter-sample peaks AND leave a little head room for lossy codec compression, for example, if you upload to YouTube or convert to MP3. Would an article on inter-sample peaks be useful, or maybe you can get the info you need via Google? Hope that helps!

  • Tommy Zai

    Hi!! As a general practice and rule, I’ve been normalizing all audio .wav files at -1dB and always choose PEAK, never RMS. In your opinion, is there a better practice/rule to follow?

    • David Post author

      If you just want the file as loud as possible with no compression artifacts that sounds fine. It’s a good idea to leave a dB spare if the file is to be compressed (for example on Youtube), this can prevent overshoots and distortion. Thanks for getting involved on my blog.

  • John Dietrich

    Great material!

    I would like to get permission to use some of your material ia a presentation I am giving to our video club this Wednesday.

    Also, how I properly cite your articles?

  • Colt Marshall

    Is there software out there or plugins for something like movie studio, audacity, etc. that has some sort of normalization feature that works well? If I don’t cut up my audio into small bits and do them one by one also manually changing gain, i have practically no results. i just want an easy was to eliminate highs or match lows/bring down master volume in my commentary files used in gameplay videos i make personally for example.

  • Prakash Wadhwani

    HI !!

    I’m a DJ & am looking to “normalize” my tracks … some tracks are really very soft. I’m not so concerned with “perceived” loudness here … just wanna ensure the track is at max volume without any clipping. Most importantly I’d like this to be performed “non-destructively” ie without any transcoding performed so as to avoid any detriment to the mp3 quality.

    I have mptrimPro … What’s your view on that ??

    I’ve also heard about mp3gain … never tried it … what’s your view on this too ?

    Any other software you’d suggest … free AND paid … I’m looking for the best in both categories.

    Thank you so much for your time.

    • David Post author

      There is already a perfect technology for DJing called ReplayGain. It analyses (a selection of) your music and attempts to make it the same volume, which is what you really want. When you say ‘max volume’ it has to be compared to something else. If you just normalize everything, like this article talks about, then some tracks will sound a totally different volume due to their peak levels. We hear in RMS, but a digital normalize function is using PEAK value.

      ReplayGain stores the volume information it generates inside your mp3 file in a tag, so it is totally lossless. There are different programs for this that work in different ways, some use RMS and some R128 (similar to RMS, but modelled on human hearing). You can analyses the music per track, per album, or for you, probably the entire DJ set. I used Foobar, but all the major DJ software has a version of this built in.

  • Ernest George

    I am not a sound engineer… Just a “listener”.. Why doesn’nt some one devise a chip or circuit that would “intercept” the audio signal of a TV set just before it hits the speakers to “normalize” the volumn… So that no matter what volumn the program or commercial puts out, you always get the same volumn outof the speaker , whatever you have set on your TV.. Seems simple enough, even though the advertisers would probably not like it, I would happily pay for such a feature on a TV..
    Thanks…E. George

  • Gilbert Perez

    Hi David, actually I’ve read this post from top to bottom and all the listed points are quite useful. Although I don’t normalize audio frequently, I will use the normalization function like you suggest.

  • Gahariet

    Hi and thanks for the informative article.
    I use Matlab to normalize several audio clips. I wanted them to be comparable in RMS. But once I made them comparable, they clipped when saving them (because the amplitude values exceed the range (-1 1). I found out the only way to save them without clipping is by increasing the bit depth to 32 (from initial 16), as to allow greater values outside the (-1 1) range. Now my question: can I save the clips in bit32 depth and when they are reproduced somewhere else they won’t clip? This is my worry. I need to process this for a perceptual experiment and I am worried that the device used will need to be of certain quality (or maybe not). Thanks!

    • David Post author

      I think the 32bit files will clip on some devices and not others, so I recommend that you make sure that none of the samples clip at 16bit. I suspect your 32bit audio is also float rather than fixed, so that may affect the clipping depending on what software is playing it. It is probably reducing bit-depth somewhere depending on your sound-card output capabilities.

      If you want accurate comparable volume, I made a video on how to use a free tool to get great results. It’s linked here with the software:

      Sometimes you will need to use a limiter on the master bus if you want to RMS volume match and still have peaks all the way up to 0dB. You can also just turn down the waveform peaks in the audio editor manually, this can be the best quality way of doing it if there are only a few.

      There is also the issue of inter-sample peaks causing clipping. You could normalize to -1dB peak max, which is what I do for anything that is later to be converted (for example YouTube). You can use free meter to measure inter-sample peaks in the ‘true peak’ reading.

      Some cheap DACS distort at 0dB regardless of inter-sample peaks, I would just default to -1dB, the loudness war is over 🙂

    • David Post author

      Could you elaborate? I am not sure what you are doing when you say “back to the Daw through cd”, and do you mean -12dB? Are you talking peak or RMS readings?

  • Geoff Lee

    “It is a destructive process. Performing any digital processing to a file is going to change it. Its bad reputation was mainly earned back in the days when digital files were all stored as 16 bit. If you turned the volume down you effectively reduced the bit depth. Your CD quality 16 bit file could end up 12 bit or less. Even turning it up with PEAK normalization caused damage.”

    Normalising applies a fixed amount of gain to the audio and that can be removed in a none destructive way. It might help watching this as a brilliant primer to the technicalities of digital audio as to why what you said about bit depth doesn’t actually matter.

    • David Post author

      I am a big fan of that video! Nothing said in the video invalidates anything in this article that I am aware of. Have a look at this:

      “Firstly, gain is processing. When you apply even a simple gain change, it expands the word length of your digital audio out to arbitrary size…. You might think digital audio is automatically accurate, but that’s far from the truth. We hear degradation in the resolution domain as flatness, cardboardy-ness, and it’s cumulative.”

  • Paul Light

    Hi David

    I use Nero 2017 when I compile and rip compilations. As you probably know Nero has it’s own normalizer built in. When there is a song within that compilation where drums (for example) aren’t present the Nero normalizer boosts this song higher than the other songs on the compilation that do have drumming (if you see what I mean).
    Is there any way that I can overcome this problem?

    • David Post author

      I have not used Nero for years so don’t know exactly what it is doing. What you need is to use ReplayGain technology. This adjusts the volume on playback losslessly rather than altering the original file. You scan your files and the meta-data is added in tags to them. You can scan per album and it should adjust individual tracks to playback at approximately the same level. This will mess with the artists intention of dynamics, so you might want to try and make your entire collection of albums match better instead. There is no one-click solution to this, but ReplayGain is where you should start looking. Please let me and the other readers know how you got on. Foobar is a fantastic music player for this job on the PC, or Deadbeef on Linux.

  • Sasha

    I use Audio Normalizer from Solar Wind. It is very fast and simple, but personally I do not have enough visualization (I would like to see the loudness graph)…and it is unfortunately not free.

    • David Post author

      Holy shit it cost 50€. I am thinking of programming my own in Javascript and hosting it here on the website. Only recently has this become technically possible to do fast and well with a modern browser… we will see, maybe I will charge a small fee to subscribe to the page for a lifetime. Got to work on my code skills!

  • Brian

    I’d never tried to publish any of my home recordings until today, when I exported a file from GarageBand in uncompressed 24 bit to SoundCloud. When I played it back it was badly distorted. I reset all the gain controls on the instruments and repeated the process. At this stage I emailed SoundCloud with my concerns about the problems I was experiencing.
    Whilst waiting for a reply I fiddled about and found the Preferences settings in GarageBand, where under “Advanced” – somewhere I’d never been before – I found “Auto Normalise”, a box that was checked by default. So of course I unchecked it, repeated the exporting and uploading process – and what do you know? No more distortion. The sound was beautiful.
    And the moral of this story is: never, ever tick the box marked “Normalisation”. Do not ever go there. Do not try to justify it with fancy diagrams, scientific data or statistics. Just leave the bloody thing alone!
    That’s my advice, anyway.

  • MLO

    Problem I’m having is playing videos (movies) through my USB media player on a TV. Volume is always too low. Have to turn up TV volume to 100% and it’s still too low. Same video plays fine volume wise on computer. To get my TV to play these videos at a volume comparable to watching normal TV, I have to convert the videos boosting the volume and then play converted result on TV. Should I boost volume on these conversions or normalize? Right now with these videos, I’m having to boost volume about +15 db to get them to play on TV at the same approx volume as watching regular TV.

    • David Post author

      Sounds like the volume is fine inside the video file, as demonstrated by the fact it plays fine on the computer. I suspect your USB media player has some funny setting on, or just has the volume down. I suggest you look into that. Maybe search forums for people who use that exact player with the same problem? Good luck!

  • S. Coach

    This would be much more useful if you would include the top and most commonly used (maybe three, or five) editing software programs that use PEAK and RMS, so that the information is not just purely academic, but is actionable. As it stands, it is not. It is merely a point of departure that requires additional independent research before any of this becomes very useful. So the article is not useless, but rates a 4 out 0f 10.

    • David Post author

      I hear what you are saying, but I think it is beyond the scope of this article. PEAK and RMS normalization mean the same thing across different software, so it is easy to find a program that can do it if you now know what to look for.

      If I were to go ahead and find all the software on all the different platforms that can perform these simple procedures it would require a new article. Loads of people read the comments, so if you have any tried and tested software to recommend please put it here in the comments and people should find it.

      The best software I have come across for all volume related functions is an extension for Reaper: . Go to extensions > loudness

  • amos spector

    Many thanks for this post. Clarifying the issues in simple words.

    However, I still need some tech advice/explanation. My audio files are mostly in WAV, and I usually use either Audacity or Audiograbber to make MP3 copies. As far as I could see, Audacity (free version) does only peak normalization. Audiograbber, on the other hand, allows average volume normalization (under advanced mode). And here are the questions: 1) what are the recommended settings for (A) use normalizing when average output is??? (seems the preset is 65)
    2) But only under these conditions: (A) Can you please state recommended settings?, (B) how can I impose a rule that refers to both average output and peak level? what if they contradict?
    3) How do you recommend to set the compression box? always compress? never compress? only when needed? (I guess “only when needed” will be the answer), and what numbers to stick into the empty boxes?

    What I plan to do is create a few audio files based on you advice (without erasing the originals), try them and use the same settings for all my files.

    Another question: Would you expect the optimal settings to be different for different genre of music? Most of my files are jazz and classical (chamber music, such as string quartets, and orchestral music (symphonies and concerti)?

    Looking forward to your reply

    • David Post author

      What is your exact goal?

      You have three separate ‘units’ of music to think about, the track, the album, and your entire collection. Normalization is going to make certain ‘units’ of music closer in volume to others. Do you want all your albums about the same volume when you listen to them? Are you aware of the lossless playback technology that does that automatically ? Maybe you are trying to do something else? If you are on PC then I highly recommend Foobar to deal with all this, it works well on Wine on Linux too.

      • amos spector

        “What is your exact goal?”
        First – thanks for your prompt reply. It’s kind of you.

        My exact goal: I use a disk on key loaded with music in two situations: (A) I have a nice audio system in my room, and I want to use it to listen to music from my DOK. This is not a major problem, as I am home, near the computer, and can easily load the DOK with the wav files. (B) I drive a lot, and listen to music (again – from a DOK) in my car. This is where I want to save space, and prefer to use smaller files, realizing that quality goes down a bit. I think my car radio is not sophisticated enough to equate volume automatically at output. I believe I need equal-volume input (i.e. – equal volume files).

        I can use MP3, and not bother you, but I understand that M4A (or AAC, whatever the differences are) is/are better and more efficient in terms of quality-space ratio. I know how to use audacity to do it, but it allows only low-quality (96 kbs) files, and I failed to follow the instructions of how to improve sound quality.

        Yes – I am aware that all pieces will be about same volume, and that’s what I want (except for classical music, in which some movements are deliberately quieter than others, but I can overcome this by converting the whole piece as one unit, rather than every movement separately).

        I hope I am clear. I also hope there’s a good solution.

        • David Post author

          The highest quality to size audio compression is opus 1.3, which is free and open source:

          Just go as high bitrate as you have the space for, and you will be getting the best possible quality. If you have more space FLAC is the standard for lossless. I don’t know if your car supports those, if not, then LAME mp3 at 320kbps would be the next best bet.

          Seems you want to use Replaygain with the option to ‘hard render’ the actual output for your car. For everything you want, Foobar is the answer! You want to find a good guide on Replaygain, really get into what it can do, and then play around with the Foobar rendering. It is very powerful and will take a bit of getting used to, but then you have the ultimate audio conversion system with a GUI.

          • amos spector

            Back from a 2-day vacation. Hence – my late reply.
            Thanks for the advice. I’ll check if my car radio understands FLAC. If not – 320 KBS is what I’ll use. Actually, this has been my “standard” for quite some time now, except that I read about M4A being better.

            And – I’ll spend the day playing around with foobar.

            Thanks again for the advice and the prompt replies.

  • Boyd Cook

    I found your article very interesting. The problem for me is that this is beyond my capability as a total novice. I have music from many different sources, some softer or louder than others. I would like to bring them all to about the same level with something that is easy to use. Also what will happen to music that I add to my collection later?

    • David Post author

      I am planning a web application to do it! It will be linked on this page, if you have any specific feature requests please be my guest. It is going to take a while as it will be epic 🙂

      • Marc

        First, thank you so much for this review, it’s the best about normalisation I found on the web!!!
        So, I have lots of MP3 track 320kbits.
        When I used it in Serato DJ, some of thems have a slim waveforms some of thems a big one, I think it’s a problem of normalisation. I used always the same way to download it, from Sidify (recording the output of Spotify). When I downloaded on Beatport, same song, same MP3 320, the waveform is huge.

        So I looking for the best software for normalize all my library. I am on MacOS Sierra.

        Thank you so much for your help !!

        • David Post author

          Thanks, very happy you like the article!

          I am writing some software which works directly in the web browser, but I don’t know how long it will take. So many people keep asking what SW to use I had to do something 🙂 Afraid I don’t know Mac, but open source software that can do it include Audacity with a GUI and FFMPEG on the command line. Also look into ReplayGain, I have talked about it quite a bit in these comments.

          • Marc

            Okey thank you for your reply.
            I found MP3 Normalization for Mac, it’s a great software. Free for single track, but you want to convert à bunch of tracks it’s better to buy the software (13€ for the life).
            I try to normalize the one song :
            – one with PEAK, it’s very very loud !!! My master is clipping but the sound is good quality.
            – RMS, i think it’s better, i try -15 LUFS, it’s kind of the results of PEAK, and -20 LUFS is the correct value, i think.

            What are u thinking about it ? Better to normalize all my library with RMS -20 LUFS (i’m deejay and it’s gonna help me a lot for have the song with the same volume)

            Thanks mate

          • David Post author

            Unfortunately it’s not as simple as choosing a level to normalize to due to the difference between RMS and Peak. The key thing to remember is the difference between limiting and normalising. With limiting you actually chop the top off the waveforms and compress the sound, with normalizing you just bring the volume up to the highest level the peak allows.

            If you have a modern highly compressed track and normalize it next to a 70’s disco track with a lot of dynamics the modern track will sound massively louder. For this reason you need to use RMS to make things actually sound the same volume, but you will still have the difference between compressed and uncompressed music sounding different. As the uncompressed has such large peaks, you are in danger of limiting them if you try to match to modern tracks. This will add distortion to the transients. You need to have software that juggles all these variables and gives you options to really get the job done. The closest thing to this is ReplayGain, see if you can learn all about how it works. It was designed to solve the exact problem you have, and to do it in a lossless way that won’t alter your music collection.

          • Marc

            Thanks for your help!!
            Yes I understand everything, i gonna use RMS.
            I have only modern track (Techno Music)

            So there will be no difference between them except with some tracks “acid techno 90’s”.

            I gonna use beaTunes it’s the same than ReplayGain for MACOS.
            All my library is in MP3 320kbits. No less, no more.
            All the WAV / FLAC are in an other library because, yes compressed / uncompressed have a really big difference when you’re playing on a big sound system.

  • tej

    Nicely written article. I am actually doing a small project basically identifying distortions in sounds which is turning to be quite a hard task. I have thought to generate spectrograms and then train a CNN for finding patterns based on good or bad. I have lots of audio files ranging from 10 seconds to 8 hours. Also audio has to be converted to mono format to generate spectrograms . I might lose some valuable information in here. Another issue is I have to split the audio file size into particular size chunks and analyze because I can’t bring an 6 hours to work on it. If you have any information regarding this I would appreciate it. Thank you.

    1) Converting audio to mono. When i do so with librosa it cannot create with stereo only mono for mel spectrogram.

    2) what size or time interval chunks i should split my audio samples to?

  • Craig Perdue

    Hi David,
    Thank you for your generosity. I have various vocalizations of a bird species from different sources. I’d be reluctant to consider trying to normalize them if it were not for the fact that doing so changes some of their spectrographic (versus waveform) properties. What then are the “true” spectrographic properties? This question leads me to consider normalization. I want to make sure I’m “comparing apples to apples”. I would be delighted to answer any questions.

    • David Post author

      I am a bit confused by the wording of your question, but I will try to answer. If you want all your bird sounds to be the same volume without changing the tone then you should normalize them. You want to use RMS or R128 rather than peak, and possibly compression depending on the sources.

      Your audio has two properties, the frequency spectrum and the volume. These both change over time. I assume your goal is to alter the volume not the frequency spectrum. If you were preparing your bird sounds for commercial use then you might want to alter the frequency spectrum of some birds too, to make them all fit together, or maybe not. This requires a professional to make a judgment on a per bird basis.

      If you backup all your files, I recommend you normalize them and listen to the results, nothing lost! If you are technically minded, these days I would use FFMPEG and some kind of helper tool like this Otherwise I would use Reaper with the SWS extension, that does it with graphics.

      I tried to check your site, but my paranoia security level browser said “Error code: SSL_ERROR_BAD_CERT_DOMAIN. Unable to communicate securely with peer: requested domain name does not match the server’s certificate.”. I think you need to look at your certificate, there is something wrong there.

  • Errol

    Ok I have read the article and comments 3 times and am still at a loss… I have a flash drive with 300 mixed (mp3 and wma) songs on it that I try to use to help me sleep. Inevitable one of the songs comes out blaring and I am awoken. Is there a tool that would analyze the 300 songs and decide what the proper levels would be to get them all at the same volume, and then go back and change the volume of them individually where needed?

    Any help would be MUCH APPRECIATED and I am not adverse to changing all to mp3 if that makes it any easier.

    • David Post author

      Yes, the technology is called ReplayGain and your software player needs to support it. You scan your collection in advance and it adds metadata to your music with information on how loud to play it. This way you don’t need to change your audio, only add metadata. You need to decide if you will scan per track or per album. This is a personal choice depending on if you want the album to retain the loudness relationships of each song. For your sleep needs you probably want to scan per track.

      If you are on a PC I recommend Foobar, it has everything you need built in. I don’t know about WMA format, if it does not work for ReplayGain then convert to FLAC. MP3 definitely works. Hope it goes well and you can sleep!

  • Dan

    Hello David,
    I am a total newbie to this field and moreover just started digging into sound classification, trying to build one ;). So, I read that normalizing audio files is highly recommended before building datasets. Am I reading the right article about normalization process of the audio files before building the classifier? If yes, what the right software/python code to normalize multiple audio files?

    Many man thanks in advance.

    • Tony

      According to my test, iDealshare VideoGo also works great to adjust volume of wav, mp3, aac, flac, mp4, avi, mov etc.
      Step 1: Click Add File button to add the media file,
      Step 2: Simply click “Setting…” -> “Advanced” tab, then go to “Audio” column, click 100% tab on the right side of volume to select volume ratio from the drop-down list.
      Step 3: Click the “Profile” drop down button to select desired output format.
      Step 4: Click Convert button to start and finishe increase the video or audio volume.
      It has both Mac and Windows version.

  • jeff

    Thanks for a great explanation, David.
    I’m fishing for normalization insights for an accent reduction project (i.e. compare your waveform of a phrase to that of a model speaker). I may actually need “abnormalization” to exaggerate the waveforms but right now I’m struggling with getting recordings made through pyaudio to look like the model recordings I make with Audacity (same mike), which are noisier and lower-resolution. Any recommendations for where I might seek solutions? Thanks!

  • Jim

    Thanks for the great post. I wonder if you can help me with this. I have some WAV files (chapters of a book I’ve recorded) where I need to increase the average amplitude to between -24 and -14 without exceeding a peak amplitude of -3. So far I haven’t found a way of disconnecting the two settings. When I raise the amplitude, the peak goes up and when I limit the peak, the average goes down.


    • David Post author

      You need compression, and maybe peak limiting. This is a big subject, but if you are after quick results you should try some free software like and see if it has any vocal presets. Check this out too:

      Your mission, should you accept it, is to make sure the audio does not sound distorted, but when you turn it down really low you can hear every word clearly. It takes years to master compression, but you will be able to get OK results with a good quality compressor, a well-chosen preset, and bringing the threshold control down to the right level.

      • Jim

        Thank you, David. With some experimentation, I think I can get close using Compressor. I combined it with Normalize to reduce the peak to -3. It’s too bad the Make-up gain for 0 dB after compressing function can’t be edited to one’s desired peak. Is the peak limiting something else available in Audacity?

        I ended up using a -36 threshold and a 6:1 ratio. Is that unusual?

        • David Post author

          Afraid I have never used Audacity so can’t help with the compressor. -36 threshold means nothing unless I have the audio file in front of me, it is totally dependant on the source material.

          For vocals a 4:1ish ratio is common, but don’t fixate on the numbers. You need to develop an ear for what it is doing, it is down to personal taste how much you want to compress. Uncompressed is how we hear in reality, compressed is ‘the sound of modern recordings’. It took me years to really hear the subtle details, don’t expect to become an expert overnight.

          I have worked on my own voice-overs for videos for ages and have found what I like. I use with:

          Algorithm: EL-4
          DRC and EL-Dyn OFF
          Classic Transition
          Knee 100% <<<< the important bit making it more like compression! Timing 100.0 RLS/ATK Ratio - 2 to 1 RLS Shape - all the way down to -1 Trans Time - all the way up to 10 Trans Shape - -0.50

          That handles the peaks too, so it is 'all in one' and saves time. You don't need this software to get great results, but it is a good compressor.

          I am going to record an audiobook myself soon, so I might make a return on this site and document some things. I have been demotivated by

  • Daniel 'World' Simmons

    So this is my issue, didn’t know what was Normalize and Amplify Gain until watching Youtube videos lately. So i have this very nice converter, popular and very efficient. It’s called NCH Switch Audio File Converter. I have over 3000+ of Mp3 Folders jam packed with music, i notice some of the files sounded low so i tried tweaking it. I usually clicks the Amplify Gain and set it to 2db and i click the Normalize which is automatically at 0db and 100.00% volume. My laptop is old and the earbuds i use are really cheap also. So i noticed after converting maybe over 1,500+ folders i started to listen some of the mp3s over again and some of the tracks sounds like when water is washing up on the shore, i never noticed it all this time. It seems as if i ruined all my Mp3 files and i am devastated. What did i do wrong, i thought that was supposed to increase the volume with no fault but i guess combing both gain and normalize was the wrong thing to do or one should of unclicked. Can you explain to me about what might have gone wrong!

  • Kent

    I have a 5 disk CD player. Different disks have different volumes. When the player changes to another disk, the sound out put level can go way up or hard to hear down. I would like a “black box” to plug in between the player and amp, or amp and speakers. Sometimes a CD can have a crescendo that is a fact of the composition. While it may be interfering with the composers creative right, it can still be jarring. When running the system to assist sleep, a sudden blast can undo and destroy the mood.
    Is there a circuit I can use to build such a device. Can I pay someone to construct one?

  • Carsos

    In my macOS Mohave iTunes music library I have around 40,000 tracks. They are either mp3 or m4a. When I create a mix for CD or flash drive, the different volume levels are really annoying. My present solution is to individually load all tracks into Twisted Wave (v. 1.22.2 – 64 bit) and then manually adjust each volume before recording the mix. I would like some recommendations for an app that would automatically normalize the track volumes for mixing, or ideally for my entire library. Any suggestions, please? I should also mention that years ago I had an app called iVolume which wasn’t perfect, but it did sort of bring the volumes closer together. However, that app is now obsolete.