Equal Loudness normalizing

It’s a common question of how to “Normalise” (Normalize) tracks so that they sound the same loudness.
There are a couple of solutions that have been developed for portable music players, namely “Replay Gain” and iTunes “Sound Check”, but to date there has been nothing for Audacity.

This is my attempt to address this omission.
The plug-in should be installed in the normal way by putting it in the Audacity plug-ins folder and restarting Audacity.

Brief instructions of how to use this effect are written at the top of the plug-in interface, but if anyone has questions, please ask here.

This is an experimental plug-in so please don’t expect perfection :wink:
My initial test have produced reasonably good results over a broad range of material, but I would be very interested to hear how others get on with it.

Achieving equal loudness automatically is quite tricky as “loudness” is a subjective measurement. This plug-in can only be developed further if there is feedback from people using it, so please give it a go and post some feedback.

20 views and 3 downloads. Has anyone tried this yet?

I think I downloaded it, but I haven’t tried it yet. Sorry, Steve.

I’d like to try this on my next dj set… One of my frequent tasks before I mix and render the sets is to try to put all tracks at about the same loudness. I usually try to amplify to the same peak value, but as one might expect that doesn’t necessarily always produce the desirable result…

I probably won’t start working on my next set until a couple of weeks from now, so my comments on this might be a bit delayed…

Likewise, I’ve downloaded it but not had a chance to check it out yet …

POL

I’m just wondering about this, isn’t A-Weighted rms the standard for loudness? Or is there not one?

A-weighted RMS is designed to measure low level noise (around 40 phon).
It is frequently used to measure other types of audio signal, but it is a poor match for “loudness”.

One of the best algorithms for loudness is that used by Replay Gain, but unfortunately it is difficult to code that in Nyquist without it being extremely slow. The algorithm used in this plug-in is a rough approximation, but hopefully close enough to be useful.

I still haven’t tried out this plugin, but I have a doubt… I’m not familiar with the Replay Gain algorithm… does this plugin apply any kind of dynamic compression to the tracks to level the loudness? Or does it apply only simple linear amplification?

There’s no compression - just simple linear amplification.

Steve,

I finally found some time to do some testing on your “Equal Loudness” effect. It works and it works on samples longer than 2 minutes. I did abandon a test on a one hour sample as the forecsat time to process was exceeding 50 minutes.

The trouble is that I can’t figure out what it’s trying to do. When I operate it on a reasonably balanced stereo pair it makes the selected sample quieter - if I deliberately unbalance one of the channels then it makes both louder.

I have sent you an email spreadsheet with the results of my tests.

WC

Thanks for the test results Peter.
I’ve found a weakness in my equal loudness algorithm that should be fairly easy to improve.
I’m rather busy with work at the moment, but will post an update as soon as I get time.

Steve, thanks very much for this. I’ve been searching high and low for a plug-in that normalises the RMS. Having read up on Replay Gain I understand it to be even better than RMS normalisation, so I’m really glad of this plug-in.
I’m not totally sure of its usage however. It seems -12 DB is the lowest possible target level. I have found -16 DB to be appropriate for most music, but then I’m still thinking in terms of RMS, where I have a good deal of experience with how much headroom to allow for certain types of audio. I don’t yet understand this plug-in well enough to know which levels to apply.
Could you elaborate on the settings? (Does headroom need to be considered as much as it does in RMS normalisation? How does it differ? Are there any generally accepted level values?)

Thanks.

This plug-in is only a first attempt and is not yet “right”. Don’t expect great results. I hope to have time to return to this plug-in some time in the future.

With the default settings the plug-in will attempt to adjust the level so that “on average” the peak level (for “normal” audio recordings) will be around -3 dB, but will be adjusted (up or down) according to the “loudness” of the track.

I’ve removed the plug-in from this topic because I don’t think that it is really up to the job.

It was originally posted just to get some initial feedback on the first prototype. As there have been over a hundred downloads there is clearly a demand for a plug-in of this type, but from the small amount of feedback I presume that most of those downloads were from people that want a fully functional plug-in rather than a desire to help with the development process.

Having done more work on this I believe that I am now much closer to producing an “Equal Loudness Normalizing” plug-in that works and will post it in this thread when ready.

This effect has now been replaced by a plug-in that calculates ReplayGain.
The plug-in is available here: https://forum.audacityteam.org/t/replaygain/22589/1