Audio export: "Don't normalise" option

• Dec 2, 2017 - 11:32
Reported version
S5 - Suggestion

When you export audio from MuseScore, the output is automatically normalised. As a result, when you listen to the audio file there is a big increase in volume. IME, when exporting audio from Digital Audio Workstations, the overall level set in the DAW is preserved.

Could an option be added to prevent normalisation of audio export?


I don't think the issue with audio normalization is just a feature request, it's a bug. The huge mismatch between the volume of in-Musescore playback and playing back an exported MP3 breaks all reasonable user expectations.

The excuse I've seen elsewhere- 'well, when you export a track you want it normalized, that's just standard practice' - is wrong. The standard is EBU R128 / ITU BS1770 normalization, and the MP3s exported by MuseScore are generally about 9dB too loud by these standards.

Hmm, my impression is that the standard you mention intended for broadcasters, not necessarily for recorded music. But it's also relatively new, and I haven't had any CD's professionally mastered since it went into effect, so I'm not familiar with how audio mastering studios are implementing it these days. Feel free to post a link to specific recommendations as to how these guidelines are to be applied to recorded music.

from My notebook. (Old standard / Analog)
I think these values are very close to EBU

for Mono / Center VU (L+R):
+3dBU == -18dBfs == 1.0V(rms)   <<-- Max.
 0dBu == -20dBfs == 0.775V(rms) <<-- Nominal
for stereo L, R (per Channel) add -3 dB for each channel

Studio/Broadcast. Levels:
European Max.: +6 dBu
USA Max      : +4 dBu

In reply to by Marc Sabatella

Though the EBU and the BS series at ITU are broadcast-related, the standard isn't just intended for broadcasts. If you've ever heard of ReplayGain, the standard does much the same thing. It defines an algorithmic measure (LUFS) which very closely matches human-perceived loudness and gives a standard loudness level for recorded works.

There is some difference of opinion about what the target level should be. The EBU standard of -23 seems to some to be overly affected by broadcast needs and too much of an adjustment for other audio. Some people recommend ReplayGain's target of -18, or something in between. But there is no real disagreement that setting a LUFS target is the proper way to normalize and that the target should be relatively low.

Peak signal normalization is NOT a reasonable way to standardize levels. It has extremely poor correlation with human-perceived loudness.

Part of the idea is that without a standard for loudness, people try to get more attention by making their material louder than everyone else's using dynamic range compression. In the world of CDs it led to the loudness war (see wikipedia). In the broadcast world, it resulted in overcompressed, overly loud channels, as well as ANNOYING COMMERCIALS WHICH ARE MUCH LOUDER THAN THE PROGRAM MATERIAL TO GRAB YOUR ATTENTION.

Though ReplayGain was created in 2001 and was seen as best practice by audio enthusiasts, rational ways of normalizing levels didn't hit the mainstream until the 2010s. The EBU standard (which is slightly superior to ReplayGain) was released in 2010. In 2011 the ITU modified their (2006-issue) standard to basically match the EBU, and it started to be more widely used in broadcasting. In 2013 Apple began requiring their Sound Check, which does LUFS normalization, for all music on their streaming service. In 2015 Youtube made LUFS loudness normalization mandatory and Spotify made it the default.

When I run r128/bs1770 gain analysis tools on MuseScore's generated MP3s, it generally tells me they are 9+ dB too loud. Since I haven't adjusted all my recordings to the EBU target, I have followed the replaygain LUFS target instead so I'm reducing levels by at least 4dB.

I wanted to add that I already opened a similar request back in 2014:
#34321: Allow to disable normalization for audio export
As this seems to come up again and again, I'll just open the discussion where to put such an option: The Audio section of the Export properties panel comes to mind (where we currently have sample rate and mp3 bit rate). Are there better places since this doesn't show up during the export process?