How to: Loudness Normalization (Manually and Automated)

[Mister Fox]

There are two ways to "Loudness Normalize" audio streams for checking submitted mixes to the "Challenge", or for Mastering Purposes:
Manually, or Automated.

In this thread, I will cover the steps how to do that.

Original Post:
http://www.kvraudio.com/forum/viewtopic ... 2#p6057492
also covered in the KVR Marks over on KVR Audio



THE MANUAL SOLUTION

The way I usually do it, is in Steinberg Cubase. But basically every host that offers stereo multi-tracks can do this.

That software can I recommend?
Pretty much every ITU-R BS.1770-x type meter out there (EBU R-128 and ATSC/85 specs are but presets of ITU-R BS.1770-x), from freeware to commercial. As long as you can set a reference level manually, or have presets like -16LUFS or -14LUFS

Take note of this thread on KVR Audio regarding available meters:
http://www.kvraudio.com/forum/viewtopic ... 6&t=476730


HOW TO DO THIS:

• have one song per channel, each track is currently at bar 1.1.1.0, so you can play all tracks at once (with speakers muted) and just analyze things
• one instance of either ITU-R BS.1770-x type meter on each channel (pre fader, no input gain compensation or HPF used), preferably set up to something that gives out LUFS (i.e. the highest available ITU-R BS.1770-x measurement preset)
• hit play and let all tracks run through their corresponding Loudness Analysis plugin
• once the playback is done, look at each plugin instance to take note of the "Integrated" loudness (ILk) and "Short Term" loudness (SLk) value
• depending on the production, now pull up/pull down the mixdown to a specific value you want to work with. i.e. if the readout on a track has been -17,6LUFS SLk (max), but you want a more similar loudness of the whole set to be -16LUFS SLk (before you do the final touches/limiting), adjust the gain of the clip by +1,4dB. If the readout has been -13,3LUFS , but I want to get to -16LUFS, then trim by -2,7dB -- IMPORTANT: this value is for "max Short Term Loudness only!
• if you trust the ILk (Integrated Loudness) value, then adjust it up/down to the particular target level you want (again, -16LUFS for example) -- IMPORTANT: this value is for "average" loudness, but is highly program material dependent (can shift in unexpected ways)

You can now move your tracks to the positions you want to have, and if you hit play, the whole "stream" (so to speak) should have a similar loudness that only needs fine tuning according to your ears. This can work with both ILk and SLk. Only that ILk uses a gate for noise bursts above -10LUFS and quiet signals below -70LUFS. ILk is more aimed at a "broadcast" stream (or radio plays, Podcasts, etc), while I personally think that SLk is more suitable for music.

!	Bonus Info
	ILk offers the benefit of "giving out" the "average value" of the stream over the course of the whole recording, while SLk only goes by "max value". To find a middle ground, utilize the Histogram Function of the metering tool (if present).

You can now add your fairy dust and use a gain->limiter chain to push up your tracks to your specific desired loudness (i.e. -14LUFS, or back down to -16LUFS), and there you have your loudness normalized/mastered record.





THE AUTOMATED SOLUTION

If you happen to use Wavelab, it's even simpler to determine the track's current loudness, which can then be adjusted to get a more "evened out" average loudness stream.

Use the Batch Function in Wavelab!

If you only want to analyze content, but not "adjust" it already:

• Create a new batch
• go into Monopass Plugins and load the Audio Analyzer
• only select "exact peak", "Integrated Loudness" and "Short Term Maximum", then select whatever output format you'd like (I like to use .csv, since it can be loaded in MS Word or Open Office)
• set the output module to "no output"
• drag and drop your songs into the data folder window
• hit the play button, the created data sheet should open automatically

If you want to "prepare" the content already, make a copy of the files, then go this route:

• Create a new batch
• go into Multipass Plugins and load the "Loudness Normalizer"
• set it up to the value you want to work with as starting point (i.e. from -23 to -16)
• set the analysis function to "short term loudness" and "exact digital peak"
• you can leave DC offset removal on, set the highest peak to -1dB and the tolerance to 0,10dB, then close the window
• drag and drop your songs into the data folder window
• set up the output module to "subfolder of the source path"
• hit the play button, the corrected files will then be put into an own subfolder

Then you can proceed as mentioned earlier.

!	Bonus Info
	There is also the program/toolset by Nugen Audio called "Audio Management Batch Processor" (or Nugen Audio AMB in short). This tool, while a bit pricey, has the benefit to actually setup an average SLk and Maximum SLk value for automated Loudness Normalization. At least if I trust the manual (haven't used that tool yet).

And this is basically loudness normalization in a nutshell.
Now you can easily check uneven loud mixes after each Challenge Round, or setup your Audio Montage for Mastering.

Hope this helps.

[Jorgeelalto]

Wow, a really detailed post. Will note and check out all this when I can, very interesting

[Mister Fox]

You're welcome

If there are remaining questions, just ask and I try to answer ASAP

[davemcisaac]

Have you checked out the MLoudnessAnalyzer from Melda Productions (part of their free bundle)? It has a "Batch Auto-Gain" function that "...lets you batch process files and make the integrated loudness match the specified Target." It works very well, but you need to save a copy of the original file because it does overwrite the file. It's a very fast way to "loudness normalize" a number of audio files at the same time.

[Mister Fox]

Definitely thanks for the bonus info, Dave. Personally, I'm aware of this metering tool, but haven't really used it since it's first iterations.

Most of the tools I use render "as new file" while leaving the original intact. There are sadly no tags in the WAV header to set both a reference and gain adjustment to handle that on-the-fly. So it's always a "destructive" edit.

[davemcisaac]

I simply right click "Copy" then "Paste" (Windows) the tracks I want to "Batch Auto-Gain" with MLoudnessAnalyzer. It only takes a couple seconds. Then I select the files and click "Batch Auto-Gain" and a few seconds later they're all done. I just experimented with the 16 entries in MC32 (so far) by auto-gaining them to -18 LUFS. It's amazing how some of the mixes are collapsing in on themselves while others shine through with all their dynamics. I don't know if you've got a way to process all the entries as they come in, but maybe it could be done prior to the judging artist downloading all the files. I'd be willing to help.

I'm going to contact Melda Productions to suggest MLoudnessAnalyzer be updated with a non-destructive process. I'm sure they are already thinking about it.

[davemcisaac]

I'm going to contact Melda Productions to suggest MLoudnessAnalyzer be updated with a non-destructive process. I'm sure they are already thinking about it.

Sadly, it's not going to happen - at least not any time soon

[OctopusOnFire]

There 's a JS plugin in reaper that normalizes at -20 LUFS iirc. There's also a script available that lets you define some parameters of the glued (consolidated) files.

[Mister Fox]

I've read of the JS scripts in Reaper at this point, but I did not take a closer look if they're of any use. The tutorial in the first post should suffice however.

Regarding MLoudnessAnalyzer and "batch processing" - I'm not surprised about the response. There cold be a "select folder where the renders should be" indeed (though I get the concept of this plugin, and it's part of the free/pro bundle), but there is no non-destructive loudness normalization possible. Read: via META DATA (there is no META tag to my knowledge for WAV to define LUFS values). It's always a destructive process.