Hello. I am a little bit confused about all the streaming formats, loudness standards and true peak stuff

1. I am producing relatively loud and bass-heavy music in Bitwig Studio and using "Mastering The Mix - Expose" to compare my master track to "industry leaders"
The final plugin on my master chain is iZotope Maximizer with -0.1db ceiling and true-peak option enabled.
After exporting to WAV or FLAC with my current project settings (24 bit/48 kHz) I get around -7.3 integrated LUFS with -0.11 db peak and -0.02 dbtp (true peak). However, when I export the same track in 24 bit/44.1kHZ or 16 bit/44.1kHz I get 0.00 db peak and +0.74 dbtp (true peak)

In 320 kbps MP3 I get even more:
48kHz: +0.77 peak, +1.10 true peak
44.1kHz: +0.55 peak, +0.78 true peak

Unfortunately, "Mastering the Mix - Expose" doesn't support OGG files (I believe Spotify converts to those), but I assume they are quite similar to MP3s. Apple Music and Youtube are going to convert to AAC which I may need to check too...

My question is: what the hell should I do with those true peaks in different formats?
As far as I understand, streaming services will take my true peak value and reduce the volume accordingly to their standards. But will they measure my original 24bit/48kHz FLAC or will they look at true peaks after converting to their formats with lower bit-depth, sample rate and different codecs?
If they take my 24bit/48kHz FLAC or WAV for measurement - I'm golden and they won't have to reduce anything. But if they look at converted ones, they might turn it down around a whole dB (which I don't want of course)

2. I also found the most recent albums or EPs by Sub Focus, Pendulum, Skrillex and Noisia (all from 2021-2023) in FLAC. I believe those were ripped from Deezer HiFi. I fed the tracks to the same analyzer and I didn't believe my eyes. Most of the tracks have 0.5-2 db true peaks and some of them have like 3-4 dbtp above zero (which is crazy to me). Are they leaving those consciously? Or could they also have 48 to 44.1kHz conversion issue after uploading?
I actually had a feeling that the recent Pendulum tracks are sounding quieter on Spotify compared to some other artists (for example in Rocktronic playlist). Of course I have "Normalize Volume" option turned off in the app. Could it mean that Spotify really turned them 1-3 db down because of the true peaks?

 

Last edited: Jul 27, 2023

No, usually the loudness is measured in LUFS and this value should be identical between the different formats. Spotify normalizes to -14 LUFS regardless of maximum True Peak. If your audio is 7 dB louder than this threshold, you don't need to care about True Peak at all, since the peaks are below 0 dB after normalization anyway.
They differ when you resample the audio and/or convert to a lossy format (like MP3), but it's also only the introduced artifacts, i.e. noise. Noise that clips is still noise. However, most streaming services recommend to reduce the master to -1 dB. If your master is louder than -14 LUFS, Spotify even recommends a headroom of 2 dB:
https://support.spotify.com/artists/article/loudness-normalization/

 

My LUFS is indeed identical between those formats, you're right

You say that I don't need to care, because Spotify normalizes to -14 LUFS (or to -11 LUFS in "Loud" mode). That's a default option, but what if a listener has normalization off. I have it turned off at all times because some tracks get butchered when a quiet part suddenly changes into a loud one. You can almost hear their "magical Spotify limiter" working with delayed "attack time"
Anyway, if normalization is turned off, I don't think the level is turned down. And in that case Spotify guys probably have to care about true peaks

I've seen that article before, but we all know that nobody is following their nonsense "-14 LUFS" rule. So I don't trust those instructions in terms of -1 or - 2 db true peaks either. I guess nobody is sticking to that. But everything above zero is still bothering me

 

If you are producing loud tracks, let peak be -1 to -2dBFS to avoid excessive ISP clipping and conversion distortion (for all lossy formats). -1dBFS is a common compromize.
That will reduce issues/distortion/clipping on the other/user end (like people playing songs on Spotify in the web browser, which doesn't have loudness normalization).
Edit: correction: peak FS, not TP.

 

Last edited: Jul 28, 2023

Ok, I get that recommendation. That's basically the same what Spotify says (minus 1-2 dbtp)
But can anyone explain what's going to happen if I don't follow that?
Let's say I'm uploading a track that is perfectly 0 db on ordinary limiter but has +1 db true peak. And I'm listening to it in the app without normalization. What will Spotify and other services do about it? Will they detect true peaks in initial file and turn it all down minus 1db even before posting the release? Or will they look only at LUFS as Olaf said, leaving all the true peaks as is? Or will the app limit/clip them only during playback? I just want to know the mechanism

Please take a track like "Sleepnet - Angel Blade" as an example and explain what's going to happen. I'm attaching the screenshot where you can see that it's 0 db peak, but 3.45 db true peak (!!!). And it's not even mp3 or ogg. It's FLAC

 

Like I said, it's usually on the user end where they can get audible distortion/clipping. It happens in the DAC (digital to analog) converter. Some converters are better than others and have some headroom tolerance. Others are more "crap". If you allow some headroom you minimize the distortion on the user's end.

Yes, ISP can be loud and clip the converter. Like you said, +3.5dB(!) even on compressed lossless like FLAC. Some guy on a crappy laptop (MoBo soundcard) might even be able to hear the distortion/clipping when listening without loudness normalization. If they playback with loudness normalization turned on they will not notice it.

 

Ok, thank you
So none of the streaming services or even distributors are going to analyze true peaks before posting my release to the store. Is that correct?
I believe it's true with Deezer HiFi because that's where those FLACs are supposedly from. But what about other services?
Can anyone confirm that they don't actually do it and only care about LUFS?

C'mon guys. I know some of you uploaded really loud stuff before

 

AFAIK Spotify analyze peak and if the material has low loudness there will be a normalization (not loudness normalization) to -1dBFS if the peak is lower than that. Say the peak is -3dBFS and the loudness is -17LUFSi, then Spotify will normalize peak to -1dBFS (and thus get -15LUFSi). It will not loudness normalize to -14LUFS, as it will go to (or beyond) 0dBFS and likely clip/distort the DAC.
Edit: correction: I mean peak. Not true peak. Edited.

 

Last edited: Jul 30, 2023

Makes sense. So they do analyze true peaks. And if they are able to turn -3dbtp to -1dbtp, then they are also able to turn +3dbtp to -1dbtp if they want. The question is - will they?
I wonder how I can test it. Probably need some kind of software to extract the OGG audio file from the Spotify app. Then I can take a FLAC of the same song from Deezer and convert it to another OGG by myself. And then compare those two OGGs. If Spotify turns something down, LUFS of the first OGG will be lower. And if it just shaves the spikes, LUFS will be the same but the true peak level will be lower than in my second file.
What do you think?

 

If you want to do what everybody does - Absolutely everybody, including at the Top40 level: use your typical limiter and then add another one that can do True Peak limiting (Ozone, Pro-L). On this one set the output ceiling to -0.1 dBFS and don't push it at all. If you use something like Ozone, set the threshold at -0.1 and if you use Pro-L set the input gain at 0. Basically you'll see almost no activity.

Export a 44.1, 16-bit dithered wav. You should confirm with a software like audacity that you have no clipping (maximum true peak stays at -0.1). At thins point you're done. Whatever extra clipping appears at conversion will not penalize your song. I routinely obtain -8 LUFS @ -0.1 dBFS TP (in the choruses) and the usual level adjust on the platforms is -4.5 dB, absolutely in line with the "competition" although when converted to mp3, there may be some places where the true peak is over 1 dB or even 2 dB.

If you export at -2 dB TP you're shooting yourself in the foot. None one at the high level (Sterling Mastering for example) does that and neither should we. Do your -7 LUFS and -0.1 db TP in the 44.1 16-bit wav and you're all set, don't overthink it.

This issue was fabricated by Ian Sheperd because he was running out of clients and wanted to draw attention on him by allegedly sharing "vital knowledge", then all youtube totorial makers jumped on the topic because they have to deliver their daily dose of meth(odology) to their consumers. This is a non-existing problem. Cheers!

 

Wow, thank you! That sounds like a plan

But can you explain why not use 24 bits and 48 kHz?
16 bit/44.1 kHz + dithering seems like a CD standard. Won't 24bit/48kHz without dithering be better if everything is done correctly? I believe Apple Music Lossless is exactly 24-bit/48kHz. Aren't we losing that option by sticking to 16 bits?

And speaking of dithering. Which one should we use? I see Triangular, Rectangular and different "POW"s in Ableton. And in Bitwg it's just "Dither" without any options

 

Last edited: Jul 27, 2023

No. LUFS (Loudness Unit Full Scale), peaks and true peaks are (three) different things. They don't "shave off the peaks" and keep the LUFS value. True Peak is also content dependant, so it varies depending on the material.

Like lxfsn points out many big houses export at -0.1 to -0.3dBFS. I generally do -0.5dBFS as a compromise. But Apple Digital Masters (previously "Mastered for iTunes") and Spotify requests/suggests -1dBFS on lower loudness and -2dBFS on loud material, "just to be on the safe side". Still, like you have seen/analyzed, ISP clipping can/will still occur.

 

Last edited: Jul 28, 2023

So you think comparing OGGs is pointless? i mean if they turn out to have even slightest differences in dbtp level, it would mean that Spotify is actually reducing levels. I only need to find a good OGG extractor

 

Do you aslo export in 16 bit / 44.1 kHz with dithering like lxfsn?

 

Top ME's from all around the world do not care about true peaks nor do they care about codec distortion.
If you master for Apple, you will need to check the conformity using AUroundtrip and afclip commands.

And about your findings about Noisia,etc : the industry does not care about true peaks.
I've assisted plenty of me's from my country and they do not like true-peak limiting and what it does on the harmonic content.
Sure you will have gurus like Ian Shepherd (who is very knowledgeable) who preach for dynamics and high PLR/PSR values but for your style of music, if you want to be on the competitive side, you cannot have -1 dBTP max peaks for a final release.

 

I think OGG has a little higher true peaks due to conversion distortion. But that's just a guess.

Yes, I dither when downsampling to 16bit and 44.1kHz. I also export 24bit and higher SR. As a rule I always (also) deliver lossless WAV the same SR as the project or the SR it came in (and 24bit).
Edit: I am a mastering engineer.

 

Last edited: Jul 30, 2023

44.1/16 offers the absolute best compatibility with anything, anywhere. many distributors do not even accept anything else. including by apple standards is considered as part of the standard:

Code:

Apple has developed its own lossless audio compression technology called Apple Lossless Audio Codec (ALAC). In addition to AAC, the entire Apple Music catalog is now also encoded using ALAC in resolutions ranging from 16-bit/44.1 kHz (CD Quality) up to 24-bit/192 kHz. -- https://support.apple.com/en-us/HT212183

As for dither type, at (y)our loudness is irrelevant imho. The lack (or the shaping) of dither at 16-bit may be audible for classical music on a tail of a very soft sound listened at loud volume through a very powerful sound system - which is so rare, makes no sense to think about it more than 10 seconds and then move on.

 

Yet I digress.. Spotify is shit, and they kept changing their standards, hence the loudness wars.

1. Spotify Streaming formats are degenerative at best.
2. Loudness Standards. Well, he who laughs at artists the loudest obviously wins.Standard Operational Bullshit
3. True Peak? That would be when EEK entered the arms race military-industrial complex funding Hellsing AI. If your music is on it, you're with it.

https://inthesetimes.com/article/spotify-military-industrial-complex-daniel-ek-prima-materia-helsing

 

 

Last edited by a moderator: Jul 28, 2023

my 2 cents:

-avoid true peaks, everyone is doing actually, if you test most of the commercial stuff nobody gives a F at true peaks levels

-export at -1 dB, this is the important parameter. Apparently commercial stuff also dont care about this in the 40% of releases (and sometime the sound very very bad)

if you want, put some test tracks on spotify and listen, also check the test tracks FROM INSTAGRAM, using the "music" function in the stories one the test files are dropped on spotiy, thats an important test for checkin your translation

once you have decided the right settings for your music, you can delete the test files and follow the same settings for ever