I have noticed there has been sporadic discussion on some member recording's loudness (or lack thereof). I figure the topic is complicated enough to deserve it’s own thread.

Disclaimer. I am not an audio engineer. I am just nerdy and a little bit obsessive.

The goals of setting proper recording level (i.e. volume) are, in the order or priority:

  1. Avoid clipping/distortion
  2. Minimize noise (from electronic equipment)
  3. Achieve appropriate loudness for playback to audience on their devices
  4. Maintain consistent dynamics between multiple tracks

To elaborate:

  1. Clipping is when the loudest part of your recording sounds like your speaker/headphone is broken. When that happens, the recording cannot be saved. Therefore, clipping should be carefully avoided in all stages of recording and post-processing. You avoid clipping by monitoring the audio level meter and reduce the gain.
  2. All electronic equipment (microphone, recorder, computer, etc) will generate their own noise (hum) and add it to the recording. Once the recorded audio file is passed onto the next stage, the noise is baked in. Any amplification will amplify that noise, too. Therefore, at every recording stage you want to make sure you signal is much louder than the noise. You do this by setting the input gain appropriately high, and of course without clipping.
  3. We all hate those obnoxious ads that are louder than the regular TV program. But a super quiet recording that forces you to turn up the volume knob can be a nuisance, too. Fortunately there is a general consensus on the appropriate loudness of audio materials for general consumption on the web. We should make our audio production conform to that.
  4. When you record multiple pieces, or when a recital is compiled from multiple players, we want to hear a soft piece sound quieter, and a bravado piece sound louder. This is especially important for piano recordings. Pop music can be normalized based on the assumption that all tracks sounds equally loud, and they won’t be far off the mark. But normalizing volume level of classical music can end up with terribly wrong results. The correct way is to figure out the gain staging parameters based on a loud piece, then leave these parameters along for all of your subsequent recordings. More on this later.

First, learn about the audio level meter.

  1. Loudest level is 0dB. Beyond that clipping happens.
  2. Common to be colored into 3 zones:
    a. Quiet and too quiet (green)
    b. Just right (yellow/orange)
    c. Caution - loud! (red)
  3. You want the meter to mostly dance in zone b
  4. But you should set level/gain based on zone c, because no clipping is priority no. 1!

Some examples:

Since most of us would be doing digital recording, the left one is what matters.



Examples of the level meter on your recorder, or recording software, and how to set level/gain for recording.

  1. Make a test recording with low gain
  2. Check the peak level
  3. Increase gain to the highest possible without peak getting above -6dB

How to make adjustment of the track’s level/loudness in Audacity. Under Effect -> Volume and Compression

  1. Amplify: just like turning volume knob up/down. You can also specify a peak level, and the amount of amplification in dB will be figured out for you. Keep “Allow clipping” unchecked.
  2. Normalize: same as amplify by setting a peak. Has additional features, such as treating stereo channels separately
  3. Loudness Normalization: loudness is similar to level (in dB) but also factoring in how human ear hear things (e.g. treble sound louder than bass). Loudness is also measured by using a moving average of a time window. Industry standard use loudness (e.g. -14dB LUFS) to specify proper audio material volume. But, loudness normalization can cause clipping! Use it very carefully, and learn how to apply compression first
  4. Compression: attenuate loud part and boost quiet part. This helps audibility in noisy environment, but hurts dynamic range. We don’t want compression for piano recording, especially for classic music.

General rule of thumb.

  1. Amplify. Or normalize. But be careful with loudness normalization.
  2. Directly off recorder, peak should be -6dB or less
  3. For final output, amplify or normalize to peak at -3dB (or less)

-3dB is my preference. But before you disagree, read on about how to (or not to) normalize multiple recorded pieces.

Maintaining dynamics between multiple recordings

You don’t want Clair de Lune to sound the same loudness as the Revolutionary Etude. But if you follow the normalize routine to every individual piece, you will end up having them sound equally loud. Arrrggghhh!

To avoid that, what you should do is to record a very loud piece and use it to figure out what your recorder level should be, and your amplification (not normalized peak) should be. And then, use the same setting for all of your recordings.

It must be noted that to do this in Audacity, amplify works, but normalize doesn’t. Normalize brings every recording to same peak which is not what we want for quiet pieces. Whereas with amplify, you can tell Audacity to always increase the loudness by a fixed amount. So how do you figure out what that fixed amount (dB) is? Using recorded piece that contains the loudest possible sound you will every make in your performance, go to amplify, set the peak to your preference (e.g. -3dB), then write down the amplification dB number. From here on, for every recording, you set this same amplification, and disregard what the new peak ends up at.

Thanks for taking the time to write this helpful guide @iternabe ! This information will help me make my next recording sound better!

Thanks iternabe, this is really helpful. I've been muddling through recording basics and have happened upon the mostly the same conclusions you have, but have still learned a lot from your thread here!

One thing I've noticed, is that indeed you audacity "good level" graph is how the theoretical ideal appears - something similar to the "average around -12db, with true peak at -6db." But when I output that, it's still tends to be on the very quiet side, especially compared to other recordings found online. The only way I've found to really fix it is to use a limiter to drop the peaks way down, and normalize closer to -3db or -1db, and honestly, it really starts looking like your "too loud" graph when it is finally comparable. I'm wondering if it's because of something I'm missing, or maybe I'm just used to the "git louder" culture of scream-casting...

    Gombessa The "too loud" graph piece has several clearly audible clipping moments. It was not my recording, so I could only make some guess. Assuming the original recorded source material has no clipping, then just like you said it will need to have some limiter applied to reduce the peak first, then amplify/normalize to bring the overall loudness up.

    But any time limiter or compressor is used, the dynamics is reduced. Even as a beginner, I find a significant portion of my time and effort polishing a piece is focusing on the dynamics. I would really want my audience to hear the end result exactly the same as I was hearing when I was playing. Therefore it makes no sense to digitally alter that dynamics just for publishing online.

    I have read about records made decades ago have better dynamics than today's. So, the "git louder" culture is indeed hard to resist. For casual listening, louder do sound better. In listening comparison of audiophile speakers, it has been shown that if the SPL (sound pressure level) at the listening position is not precisely calibrated to be equal for all test speakers, then the slightly louder speaker will have a better chance of being perceived as sounding better, even when it is not.

      Also ... the techniques presented are a good ... actually excellent rule of thumb. So then, having that excellent knowledge then allows the users to amend etc according to their desired result(s).

      Gombessa on another forum I went to great lengths to describe to you how to use a limiter. Seems like you learned the lesson well; the second wave does look good. If you zoom way in and the tops of your wave forms show flattened peaks then you could ease up on your threshold.

        7 days later

        @internabe Yes, there were several comments about my recordings all being too quiet... I think it's because I'm so used to playing with the volume very low, so as not to inflict noise on my husband, that even just turning it up slightly sounded super loud to me. I will definitely have the volume much higher for any future recordings, but I don't have a sound meter so I can't really follow your advice.
        I always feel puzzled when I think about stereo v mono recording/output. On my basic Casio keyboard there are two speakers, one each side, and I had thought that the base notes would come from the left side and the treble notes from the right side, but when I bent my head to listen I discovered that both speakers emitted exactly the same sound - a sort of mono effect, really. I recorded using my Chrome laptop which just has an inbuilt microphone. I wonder if I would have got better sound quality using a lead to connect the two devices - does anyone know? Also, if others play back my recording on a high quality device, would they be able to get it to reproduce the sound in true stereo - bass from one speaker, treble from the other? Or would that not be possible due to the basic way I recorded it? It's all a mystery to me.

        It's about to get even more confusing, as I now have a digital piano. 🙂 To me the sound quality seems much better, there is more depth to the sound, but I haven't checked whether the two speakers divide the sound output between bass and treble. The speakers are under the keyboard so it will be a struggle for me to get underneath to listen, while tapping a few keys. I'll do it at some point, but my knees aren't great so I'm not in a hurry to satisfy my curiosity on this point.

        "Don't let's ask for the moon, we have the stars." (Final line from Now,Voyager, 1942)

          Nightowl On my basic Casio keyboard there are two speakers, one each side, and I had thought that the base notes would come from the left side and the treble notes from the right side, but when I bent my head to listen I discovered that both speakers emitted exactly the same sound - a sort of mono effect, really. I recorded using my Chrome laptop which just has an inbuilt microphone. I wonder if I would have got better sound quality using a lead to connect the two devices - does anyone know?

          Excellent post nightowl. And excellent question. That is great how you listened to both speakers - and that you sense the sound content coming from each one is the same. If your casio keyboard has left and right channel audio outputs, then taking those outputs and feeding them both into a stereo recording system will certainly allow you to use software such as Audacity (which is free) to see graphically any differences between the left channel audio and the right channel audio content. The software will load your recording - such as .wav file.

          You can also get the model of your casio keyboard to look online to get the specifications, which will indicate whether it has left and right channel audio output sockets or not.

            Nightowl That "too quiet" example I posted is not you. So you are definitely not alone 😉. Although the ideal is to make the recording's loudness mirror the performance, there is some leeway for special situations. Just like in photography where you can make a dark scene brighter by increasing exposure, you can make a quietly played piece record louder by increasing "gain".

            The gain may sometimes be called sensitivity, or even just volume on the microphone setting screen of your chromebook. It will be just like a volume slider bar. Try increase it to see if your recording end up louder.

            The built-in microphone of your chromebook is likely to be very low quality. You should be able to connect an external USB microphone. Even a very inexpensive one would bring big improvement.

            Most digital piano can output audio directly to PC/Mac, but I am not sure if it's the same for chromebook.

                Thank you for the summary and recommendations.

                "...Increase gain to the highest possible without peak getting above -6dB..."

                This is exactly what I did recently using a simple Zoom H1N recorder (all limiting/auto leveling disabled). Oddly perhaps, after experimenting I seemed to get the best sound keeping the lid (of the Yamaha C2) flat, with the recorder placed on top of it and set at a slight angle towards the upper part of the keyboard. (I'm pretty sure this violates every known rule of good piano recording 😆.) I set the level of the recording device as you indicate above while playing the loudest portion of the music.
                I then transferred the recordings to a laptop and used Amadeus Pro for editing and a bit of "presence", since the originals were recorded in a living room. I'm pretty happy with the results (for a home demonstration recording):

                  Then there's the concept of "loudness". Decades ago when I worked for a television station the common concern was viewers' bitching that the commercials were louder than the programs. This would be true even if our techs diligently used their VU meters and kept the levels consistent for all sources.
                  The reason was that the commercials used a lot of audio compression.
                  A company (Dorrough) sold "loudness meters" to try to help broadcasters keep the perceived loudness level consistent.

                  Speech/dialogue has a lot of natural gaps, which decreases our sense of loudness.
                  A performance of a Chopin nocturne may have places where the VU meter registers 0 dB, but will sound much less loud than something from a metal band played at 0 dB.
                  Most VSTs have a setting for dynamic range. For classical playing I tend to set that higher than the factory default. But that also decreases the overall sense of loudness because it increases the spread between the loudest and softest passages.
                  OTOH, I've seen on Phil Best's youtube channel an explanation of what compression he (sometimes?) applies to his recordings before uploading them.
                  And driving in a car with moderate noise, having large dynamic range rather than something that's rather compressed can be a frustrating experience.
                  I think it's as much a psychoacoustic challenge as a technical one.

                  Make a joyful noise...
                  Jane - expert on nothing with opinions on everything.

                  True. Because the audio content/power - and at the various frequencies - is time dependent, we can probably only get some idea of the power when averaged over some chosen amount of time, and then the amplification could be controlled a bit - maybe automatically, depending on the average power etc. On the other hand - for piano music, they usually don't want to keep automatically changing the sound level or change compression levels for any part of the piano music playing, to avoid altering the audio levels, as the piano player wants to have control over the audio level, with no external influence.

                  pseudonym58 I liked you’re recording it had a nice tonal quality like the the instrument as a whole was being captured.

                    Interesting discussion.

                    An acquaintance of mine did some solo piano recording not that long ago and he mentioned that he did use a bit of certain tools, compression being one of them. It's not necessarily an on/off kind of thing and apparently a person can use "a little", kinda like reverb or some other effects, where too much is a bad thing and yet a little can enhance certain things a bit.

                    Dunno, I'm not a recording expert by any means, just an observer hopin' to learn a little somethin' every once in a while. 😉

                      In terms of reproduction, it's also interesting to note that many people listen to music through Bluetooth headsets, and these in general incorporate a certain amount of compression, as I understand it (although the amount probably varies greatly depending on the Bluetooth version, codec and quality of the devices used).