[chillysalsa]

Came across this really interesting lecture that explains what happens when the audio loudness gets compressed on uploading to a streaming services (note, that's not data compression, but the audio amplitude loudness compression).

It's worth watching the whole video if you want more background, but here is the time stamp with the key explanation. He demonstrates isolating the effect of compression, by compressing one song, and then inverting it, and adding it to the original, ie: any remaining audible sound is the compression garbage noise we're all listening to on streaming!

 

[danadam]

explains what happens when the audio loudness gets compressed on uploading to a streaming services

It should be "before uploading". "On uploading" makes it sound as if it was the streaming services that were doing the compression. They are not (*).

*) Well, some of them may if the track is really dynamic but the presentation is not about that.

any remaining audible sound is the compression garbage noise we're all listening to on streaming!

Not only on streaming. On CDs and anywhere else too. The difference is that streaming services actually give an incentive to not do the (excessive) compression. That's what the presentation is about.

 

[chillysalsa]

To me it sounds like if your track is exceeding the -14 db LUFS value, the streaming service will compress it for, so that they have all their content playing close to the same loudness, and we control the 'loudness war' by removing the incentive to increase your track loudness.

So, I guess in a way they are trying to preserve some standard level of dynamic range, which is good compared to the over-compressed sound on the radio.c

 

[danadam]

If a track is louder than the streaming service target, then the track's volume will be reduced to meet the target. This is not a compression, just a volume change.

If a track is quieter than the streaming service target, then:
* it may be played as is
* or it may be made louder with addition of some compression to prevent clipping.

AFAIK youtube does the first, spotify does the second. And in this second case the compression is applied only to the peaks that would exceed 0 dBFS, so usually only a small part of the track is affected.

 

[chillysalsa]

Well that is good to know. Right, so it seems at least for spotify, they claim:

Playback levels are not adjusted when transcoding tracks. Tracks are delivered to the app with their original volume levels, and positive/negative gain compensation is only applied to a track while it’s playing. This gives users the option to adjust the Loudness Normalization if they want to.

• Negative gain is applied to louder masters so the loudness level is at ca - 14 dB LUFS. This process only decreases the volume in comparison to the master; no additional distortion occurs.
• Positive gain is applied to softer masters so that the loudness level is at ca - 14 dB LUFS. A limiter is also applied, set to engage at -1 dB (sample values), with a 5 ms attack time and a 100 ms decay time. This will prevent any distortion or clipping from soft but dynamic tracks.

So, as you say, it seems only if it is too quiet, they will apply a bit of limiting to keep it from clipping, so there will be some minor possible temporal distortion for some tracks that are quieter. It's possible you could hear that being different when there are transients, when you compare to the original source on CD or original master.

 

[bigshot]

The quality of the sound on streaming all depends on the master the record label gives the streaming service. Some streaming services, like Apple have delivery specifications, but the way the label has mastered it makes the most difference in perceived sound quality. It's really no different for CDs. You can often buy the same album in several different masterings. Serious collectors discuss which release is the one that has the best sound on music forums. Sometimes they know what they're talking about (no different than any internet forum).