I think it is a simple psychoaccoustic model - it is an analyzer that will spit out a normalized averaged loudness measure. The louder you push your sound, the higher it will be. Sound loudness is very complex, and the model takes different things into account that can’t be averaged by simple meters. To go for it, just put the meter on the output, view it as magic box that you can read the value from, then crank your mix, push the limiter/maximizer until you get the desired loudness. If you want to go fancy, learn about proper mastering in general, it is an art in itself to get well-formed sound. If you need it quick and dirty, just use a maximizer and boost your sound into it until it is loud enough. For soundcloud -14 is their optimum, but you can also go higher and then soundcloud will reduce your levels, and you have a sound with less dynamics…which might sound suboptimal, but is actually what some people want for when having it played back loud. I think you should just playback/monitor your song through the meter plugin, and see what the average is during the louder passages of the song. That is the value you want to master for the target.
That’s strange, I usually boost up to -10, but didn’t realize a reduction of volume. Maybe I have to compare it more accurately. What I am surely know is that it doesn’t really work on SoundCloud, very likely that the next played Song will be super loud. Their calculations obviously suck. So I wouldn’t care at all 
Also if the goal is that all songs have have kind of equal loudness, why care at all? If it is too quiet, it should be boosted, if it is too loud, it should be reduced. In any case, you should upload a wav / raw file.
Their calculations are probably fine. More likely that what you’re hearing is a difference in the mixes. The distribution of energy, if you will.
For example, if you have a lot of excessive low end rumbling around in your track you’ll be surpassing the -14LUFS threshold quite easily. It may not be something you can hear but it does eat up energy and headroom nonetheless. But once you clean that up, make it tight, suddenly you can push your entire mix up by a couple of dB before you hit that -14LUFS standard.
The -14LUFS doesn’t prevent you from making something very loud sounding. You just need to take good care of in what frequency range you put that energy. Loudness starts with the composition and arrangement.
Yes sure, but my point was this: The goal of the algorithm on Soundcloud is to match the perceived volume of the tracks. Yet it completely fails, you still are getting very loud and then quiet songs in your song queue. As long as those algorithms do not also look into “density” of a song, so the instruments in a song which hide behind th eloudest peaking instrument, it will fail. So I personally do not care about it, at least on soundcloud.
Thank you all for the perspectives I do appreciate the replies from people intimately familiar with this topic. My main issue is just for example on New Years Eve at my friends place I wanted to play a few of my songs at a party, and they sounded weirdly quiet compared to the rest of the music played at the party even when we put my friends speaker system to max, and for sure it’s a result of normalization on soundcloud due to this LUFS system.
It is true as above posted by Eretsua that I do have a lot of excessive low end bass or rumbling at times in some of my tracks and that may be related. Honestly speaking I bought a CD burner recently and I am more inclined to just burn music onto CDs where it doesn’t get interfered with with this LUFS or EBU R 128 shenanigans.
Of course another option is that I ‘git gud’ but my forte has always been music composition and so forth (have a classical music background), and I am weak on the ‘sound production’ side.
I will also investigate the notion of ‘headroom’ raised above. I have been experimenting with these LUFS meters but they are a little above my paygrade, I guess I have to struggle to figure it out if I want to keep using soundcloud. I do fully appreciate that mastering is an artform in itself, just this particular problem causes me great consternation.
Thank you once again for all the informative replies above.
That or find someone to collaborate with who is more into that side of things. Don’t forget that a lot of the famous and successful artists have scores of people working with them. Helping with, or actually doing the engineering, mixing and especially mastering. Sending it off to a professional in a different room who does their magic. You’ll be surprised at what a difference that can make. To actually have your tracks properly mastered.
Totally, originally my reply included 'that’s one of the gripes I have with the LUFS system. It doesn’t actually tell you how loud something is going to be perceived", but it got scrapped while typing. So I’m with you on that.
Just being curious, from this page:
How is LUFS actually calculated?
In short, the audio is run through a K-weighting filter, which boosts high end and rolls off low end to compensate for the frequency response of human perception. The audio is also gated before the analysis; this prevents people from adding a bunch of silence to their tracks, bringing the program loudness down and thus letting the rest of the song be super loud (for example).
Finally, an average signal value is measured from the output of these processes. Loudness meters often measure both momentary loudness (the last 400 ms or so), short term loudness (2-3 seconds or so) and program loudness (the whole song). The last one is especially important, as it determines how the song might be normalised on streaming services like Spotify.
So here you could already extract ways to circumvent this algorithm to be the loudest in the hood, similar like:
- Do always add a range limit to your sidechain ducking, make the ducking as long as possible, stay above the algorithmic gating
- In your very maximized track, add a lot of breaks with quieter parts, just automate master volume or so
- Use very short drums
- etc…
Some LUFS meter code. And here the paper.
According to their own website you’re right and it’s -14 LUFS. So all the LUFS tables seem to be not always right. Here’s another one with a few more values:
Sounds like you’ve got some issues in terms of mastering. What kind of music do you procuce? Do you use a maximizer? Back in the days I had the same problem and my music was quite soft in comparison with professional productions. So I learned how to master properly. I think the key for some more loudness is to use a maximizer and push the volume below the threshold. Of course the mix needs to be quite right, too. If you don’t have any experience how to I would recommend to have a look at the mixing thread. Get rid of the dirt (unnecessary low frequencies), check the frequency spectrum and then do your master. If you do so you could ignore the LUFS in terms of Soundcloud. I never checked the LUFS in the past and I never had problems in terms of loudness, regardless of the platform (Soundcloud, Youtube, whatever).
It’s free. 
Thank you for your tips TNT, you have been helpful in this thread and I’m pretty sure also questions I asked in the past.
In answer to your questions: What kind of music do I produce - I am spread pretty wide, I guess hardcore (as in gabber), hardstyle, trance, synth, classical, tekstyle, Influences of DnB, trap, Often I have 808 basses that might push some kind of mastering limits, or also distorted kick drums. But also I have made quite a bit of ‘no drums’ music which often I find becomes very quiet also. I use a lot of instruments from the bbc orchestra plugin or other acoustic instruments from LABS Spitfire Audio.
That was the last track I uploaded, perhaps it typifies the kind of stuff I often make.
Honestly speaking, I do not use a maximizer generally. I would have to say that I do not know its use exactly, although I frequently use a gainer or a compressor on instruments, and many other common effects such as wave packets on filters, or EQ.
Could you tell me in a nutshell what a maximixer does? Somehow it is not an effect I have much history with. What do you mean by “push the volume below the threshold”?
EDIT: And I would just mention I feel the volume is not too bad on the above track I linked, but on other tracks I uploaded around the same period, it was substantially lowered. I do feel like they killed the 808s on the above track though.
Ok, so you’re producing mainly electronic music, which means you should use a maximizer.
At least I would recommend to do so.
Sure. On one hand a maximizer simply maximizes the volume of the signal, on the other hand a maximizer limits the signal just like a limiter. As you might know everything above 0dB most likely causes clipping and artifacts, especially if you’re planning to upload on Soundcloud or elsewhere, and we don’t want that. So you need something to limit the signal to prevent clipping and artifacts. With a maximizer you can increase the volume of the softer parts of the signal until it reaches 0dB, therefore increase “Boost” if you’re using the native maximizer:
In the end you’re simply increasing the volume of the signal without getting above 0dB. Just use this default settings and increase “Boost”, that’s all you need to do besides putting the maximizer in your effect chain in the master track. I would also recommend to learn about limiters and clippers, too. There are plenty of videos on Youtube for example, check this.
The threshold is your upper limit of the signal, adjusted by yourself. If the signal reaches the threshold, the maximizer (or whatever device) starts to work, in this case the signal is getting limited. When I’m talking about pushing the volume below the threshold, I’m talking about increasing the volume of the softer parts until it reaches 0dB, just like mentioned above.
One (perhaps obvious) caveat with using a limiter or maximizer is to listen carefully to the sound to make sure that it’s not breaking up/distorting/changing the character of the sound in a way you don’t like. Noisy sounds can take a lot more limiting than sine-wave based sounds without losing much quality. Kick drums can definitely get flabby sounding if over limited/maximized
This was something it took me a while to appreciate.
Also, while steaming services may tone down overly-loud tracks, they probably don’t boost quiet mixes. There’s no quality control for what gets uploaded to Soundcloud so variations in perceived loudness can have multiple causes.
SC don’t boost the levels on uploads unless you use their built in Dolby mastering service. It adjusts the levels so if it’s too loud it will lower the volume, but if it’s too low it will boost the volume and i think it makes it -14LUFS.
Can anyone tell me in plain English what these LUFS algorithms penalise, resulting in an overall reduction of the volume of the entire track? One comment in this discussion by Eretsua mentioned that it penalises ‘low rumbling sounds’. What is the other naughty behaviour that it aims to deter? Again, in laymens’ terms if you can.
For example, I routinely listen to tracks of the same genres I compose, that are to my ears, much much louder than anything I can upload. How do they do it? Inquiring minds want to know.
Thank you for any replies.
Sausage Fattener (Saturation Effects):
Any LUFS meter with presets for streaming sites:
Hth
Anything that’s “too loud”. It doesn’t matter which part of the frequency spectrum. But you have to keep in mind that the integrated LUFS is an average value of the whole song, which means there can also be pretty loud parts of the song. LUFS metering is different than limiting/clipping/maximizing (limiting/cutting the peaks), it’s more like a “volume knob”. But you have to control the volume via limiters/maximizers by creating a good overall balance without too strong peaks.
As I said before, that’s a matter of mixing. Make sure your mix is good before you start measuring and adjust to the desired LUFS value. Use filters in every track to get rid of the dirt, use conpressors in every track to tame peaks, use a maximizer in your master to increase the volume just as described above and to prevent clipping and artifacts and if necessary use a limiter (TP limiter is recommended), too. The better balanced your mix is the louder you can make it while mastering.
Gaining more LUFS loudness is simply a matter of mixing and mastering well. You can adjust the peaks to be right, you can compress the bass hardly so it uses the maximum range, you can overdrive all kinds of sounds to make them more present and loud, and you can limit/saturate your sound to the max to push all the info into it like fitting them into a crammed space and then even blowing them up to fill it to the maximum. So it is possible to get those -8 LUFS mixes with mastering, or even like -4 for extreme mastering of harsh/hard music.
The simple solution to see is to put a maximizer (is a limiter…) on the master and then crank the boost until the LUFS average reads the desired amount during loud passages of your tune. Then you can see how the sound is squished together and what kind of loudness it yields. If your mastering before the limiter lacks, you might have problems to gain the desired loudness and you might experience distortions in your sound, as well. To get rid of the distortions, you have to fix it before the limiter, i.e. multiband compress, eq, getting the mix right including properly fixed transients and compression of all instruments etc.
Regarding the mix…I think that LUFS heavily penalizes if you use too much sub. It will also penalize if your sound is too harsh or too dull. So you must get these right in the mix. I like to aim for a resulting peak sound spectrum like a pink noise curve which is level below the bass range like < 200-300 hz, and the deep sub below 100 hz I also let roll off so that the freqs around 40hz are lowered in level in comparison to the other sub/bass. In the past I also used too much sub…sounds good on headphones or with studio monitors, but use with crappy equipment that has muddy bass then the sub will drown everything in rumble and make it sound disgusting. If you want your sub to be more prominent, rather try to add harmonics or saturate the mix so the other freqs react to the sub.
Rounded squares or squarish sines work quite nicely for loud even subs that have some decent harmonic content. Sometimes more consistent and easier to tame than a distorted or pure sine, for me at least
Could it be that you have some peak somewhere in your song, and that the algorithm that Soundcloud uses, takes that peak as your overall volume before normalising the file? That way, your track will be quieter then other tracks. Peaks won’t always show up in your waveform btw.
Cranking the volume up of your track before running it through a limiter would fix that.
It’s short and maybe a bit too short if you’re inexperienced with this stuff. But Dan Worrall’s core argument “just make it sound good” and taking proper care of the mid range stand firm.
