I am doing a standup routine that I will upload to Youtube. So the goal is to follow the -14 LUFS recommendation. I want my sound to be as loud as possible within that limitation. It is only my voice in the recording.
Naively I used the Loudness Normalisation tool in Audacity ( after I had cleaned up the breathing sounds )
I set loudness normalisation to -14 LUFS.
I have the Youlean loudness meter installed and I am looking at the integrated LUFS.
What I can see from the Youlean Loudness Meter is that the integrated LUFS is rather -22 LUFS.
In the waveform I can see that I have some peaks above -15 dB. I run the Compression tool with Level set to -15 dB and I use a 4:1 compression. I have the compression set to automatic makeup gain to the zero level.
This way I get closer to the -14 LUFS.
Now I am wondering. What is the difference between Loudness Normalisation and Normalisation in Audacity? I mean normalisation is pretty much amplifying the whole signal with the same amount ( as I understand it ). I mean it would be easier to use True Peak normalisation and just set that to -1db. It would be easier because I got clipping with Loudness Normalisation in this particular instance. In practice I do the tweaking with the compressor to get close to -14 LUFS.
This whole journey started as I upload clips to Youtube but they are reported as being -25 LUFS by Youtube. I want to get closer to the -14 LUFS level.
With Audacity’s “Loudness Normalization” effect, ensure that “perceived loudness” is selected, then the selected audio will be correctly normalized to the specified level.
Note: For the exported audio file to have the same loudness level, it is essential that:
The project has only one audio track
The “Pan” and “Gain” sliders on the left end of the track are both centred (“Pan = Centre” and “Gain = 0 dB”)
Note also that there are different methods for measuring the loudness of mono tracks. Audacity’s default is to treat mono tracks the same as a stereo track that has the same audio in both channels (which is what you hear on playback).
So the goal is to follow the -14 LUFS recommendation
Are you sure that’s their recommendation, or is that their target for their own loudness normalization? Does that apply to spoken word?
Now I am wondering. What is the difference between Loudness Normalisation and Normalisation in Audacity?
Regular normalization is based on the peaks. Loudness normalization targets perceived loudness. The peak levels don’t correlate well with loudness, and if you’re not careful with loudness-normalization you can push your peaks into clipping.
I mean normalisation is pretty much amplifying the whole signal with the same amount ( as I understand it ).
Yes. Regular normalization and loudness normalization both work that way.
I mean it would be easier to use True Peak normalisation and just set that to -1db.
If you want to go as loud as possible (linearly) you can normalize to 0dB. It doesn’t hurt to normalize (to 0dB) as the last step. FYI - The Amplify effect will normalize to 0dB if you leave the default. Don’t worry about the “true peak”. The “true peak” is an approximation of the re-constructed analog peak (in-between the digital samples). Audacity will normalize the actual digital data.
It will help to level-out the sound (with the envelope tool or the leveler effect) and you might try some limiting. The loud parts/peaks of your performance are the limiting factor. Compression also evens-out the volume. Limiting is a kind of compression and it’s a good way of making your recording “louder” (as long as you amplify/normalize or apply make-up gain.) But, you may want to leave some dynamic expression in your performance… It’s up to you, but it might get boring if it’s all constantly-loud.
Over-processing (or the “wrong” processing) can make it worse! So don’t get too hung-up on “loudness”.
Note that YouTube applies their own loudness normalization and the viewer/listener has a volume control so you don’t have total control.
Thanks for answers.I am still struggling with Audacity.
I did a two tests today and exported to Youtube.
In Davinci Resolve I set Target Loudness Level to -14 LUFS for the Project ( Loudness is measured with BS.1770-1 standard). The Loudness Meter shows an integrated LUFS of -1.9 (relative to the -14 LUFS target) .
I upload the video to Youtube. In Stats for Nerds I get the information that the content loudness is -1.9db. Volume is 100/100 ( meaning Youtube has not touched it, since it is below the -14 LUFS border ).
So I have a pretty good preview in Davinci Resolve on the room I have before Youtube will trim it down.
I am trying to set up the same thing in Audacity. I am using the Youlean Loudness Meter.
But what the Youlean Loudness Meter shows me in Audacity is not the same compared to the Loudness Meter in Davinci Resolve.
I am totally confused by this.
But what the Youlean Loudness Meter shows me in Audacity is not the same compared to the Loudness Meter in Davinci Resolve.
I am totally confused by this.
I don’t know why there’s a difference. LUFS is well-documented and standardized. Maybe it’s something related to mono/stereo? The difference/error is probably constant.
If I have some time later I’ll try some experiments…
Volume is 100/100 ( meaning Youtube has not touched it, since it is below the -14 LUFS border ).
If your file is normalized it can’t be boosted without clipping and that’s probably why they didn’t touch it…
If it doesn’t sound unusually-quiet compared to everything else on YouTube I wouldn’t worry about it… I wouldn’t worry about 2dB. Just for reference the audiobook standards allow for 5dB loudness range (-23 to -18 dB RMS).
The project only have one audio track.( it is an XLR microphone so I have one mono track ). I have not touched the Pan and Gain sliders on the left(both on zero).
I selected “perceived loudness” and that is the only option that lets me enter a LUFS value.
I’m also getting inconsistent results. With mono files I’m getting a 3dB error/difference. I Loudness Normalized a mono file to -23dB. dpMeter4, Youlean, and GoldWave application all report -26dB.
When I tried the same experiment with a stereo file I got:
dPMeter 4 = -24.8
Youlean = -24.3dB
GoldWave = -23dB
Looking through the forum, I’ve found a couple of other topics that report that the Youlean meter, when run in Audacity, gives incorrect results. I’m not able to test this as I don’t have the plug-in and I’m not on Windows.
An easy way to test is to use Audacity’s Loudness Normalization and then see if Youlean agrees.
Exporting from Audacity as 32-bit WAV with produce a file that is a perfect copy of the audio in the Audacity project (provided that track Pan and track Gain are centred and the Project Rate matches the track sample rate). The exported file can then be tested in other software to compare results.
I have a high degree of confidence that Audacity’s Loudness Normalization is correct, and would expect most other tools to agree with it.
You can use the Youlean scanner instead of Loudness Normalization. For example, if you’re reading -16dB and you’re shooting for -14dB simply Amplify by +2dB. You may have to allow clipping but that’s no different from the Loudness Normalization effect which doesn’t check for clipping. You can compress or limit to bring the peaks down out of clipping* but that will affect the LUFS to some extent so you may have to iterate through the process one or more times to “zoom-in” on -14dB.
The floating-point data isn’t actually clipped until you export to regular integer WAV or other format that’s limited to 0dB.
( I have tried the Free YouLean plugin for Audacity on both Mac and Windows and the metering is off by at least a few LUFS. This seems to be an issue with YouLean tool and not with Audacity itself. Probably something I should talk with Youlean about )
I transformed the mono track into a stereo track in Audacity. I did a loudness normalisation on the stereo track to -14 LUFS. I got some clipping but ignored that for now, as this is a test. I exported the audio into MP3 and I got -14.05 integrated LUFS in the Youlean Online tool. I got about -14.02 LUFS in Loudness Meter in Davinci Resolve. I got Youtube stats for Nerds to be +0.2 db ( which is pretty much on target with the -14 LUFS).
Thanks for all the help in clarifying! I have a feature request for Audacity and this is Integrated Loudness Metering in LUFS
