You must log in or # to comment.
Here’s an archive link, in case anyone, like me, refuses to read anything on Substack due to their affinity for Nazis:
What use is the metadata? Or is that stuff like the album covers etc?
I’m also a bit concerned about this:
Second, there’s an obsession with audiophile-grade quality (lossless FLAC, etc.) that inflates file sizes, making it impossible to maintain a complete archive of all music ever produced.
Does that mean that this spotify dump is a bunch of 64kbs mp3s, or worse some kind of lossy spotify transcode?
Spotify streams all music at 160kbps OGG for free users by default, so that’s what this archive is dumped at - the original Spotify content, no transcode. The only difference is they re-encoded all the songs with a ‘popularity’ of zero at a lower bitrate, because that saved an enormous amount of data for all the AI crap pumped into Spotify that nobody listens to.
Side note - it would probably not be possible to do a dump as a paid used (as they would notice a user account is being abused, and ban it), but paid accounts go up to 320kbps OGG and some content is also available lossless (as FLAC).
Anyway, 99%+ of people can’t consistently tell the difference between a 160kbps OGG and lossless, because of limitations in either their equipment, training, ears, or a combination thereof. This has been blind tested many times and the audiophiles that ‘swear they can tell’ are always proven wrong, they then usually blame the equipment or test. There’s tests you can run yourself too, eg here: https://abx.digitalfeed.net/list.html
Heh, like wine tasting.
Anyway, 99%+ of people can’t consistently tell the difference between a 160kbps OGG and lossless, because of limitations in either their equipment, training, ears, or a combination thereof. This has been blind tested many times and the audiophiles that ‘swear they can tell’ are always proven wrong, they then usually blame the equipment or test. There’s tests you can run yourself too, eg here: https://abx.digitalfeed.net/list.html
Ooohhh I did that test when I got a new speaker / amp setup at my PC and as a musician I thought “I got this”. Plus I was trying to decide if Tidal was worth upgrading to from Spotify.
I did slightly better than average. Like just slightly. I might have the results somewhere.
I ended up doing Tidal’s free trial. I couldn’t tell a difference. Went back to Spotify. (though now my group of people are on an Apple Music family plan).
I tried Tidal and could not go back to Spotify.
I guess it also depends on what type of music you are listening to. Simple FM pop or hip-hop works fine compressed. Rock, classical, melodic or more complex music gets the high range completely smashed by artifacts.
They definitely don’t see paid accounts using too much. My library is mostly built from Spotify rips using onthespot. 400k songs.
Everything is broken now and a few patches are up, waiting for it to update and keep ripping.
1000% agree most people can’t tell the difference in the bitrates, I rip mine at 320 in mp3. I can’t tell the difference in flac even with decent headphones, not worth the storage for the extra file size, I use Bluetooth headphones anyway, don’t care about every single frequency. I did notice some of my super old 120kbps mp3s from the 90s were so shit quality though, redownloaded those.
I try to keep lossless on my server for the fact that new algorithms come out all the time. I don’t want to be stuck with a 160k mp3 when a better algorithm comes out or if I need to stream just a little lower than that. I’d rather have lossless quality that can be converted at any time to whatever I need, even though I mostly have it set to 160k for listening if I’m streaming away from home. My work internet and cell service can get really terrible, and being able to buffer 10 songs when I get a few minutes of service is a godsend while not getting stuck with low quality, several times converted files, or keeping multiple bitrate versions of the same song.
Sure but if the purpose of the dump is musical archive, then the music should be stored in an archival format, not a lossy one.
They are archiving what they could get from Spotify and in this case what they could get was encoded in a lossy format to begin with.
It doesn’t get any “more” lossy over time just being stored as an .ogg
- For
0, we got close to all tracks on the platform. The quality is the original OGG Vorbis at 160kbit/s. Metadata was added without reencoding the audio (and an archive of diff files is available to reconstruct the original files from Spotify, as well as a metadata file with original hashes and checksums).
- For
popularity=0, we got files representing about half the number of listens (either original or a copy with the same ISRC). The audio is reencoded to [OGG Opus](https://en.wikipedia.org/wiki/Opus_/(audio_format) at 75kbit/s — sounding the same to most people, but noticeable to an expert.
- For
Modern pop music is already compressed to hell, FLAC is a waste of bandwidth unless you’re encoding orchestral pieces.
I think you’re confusing two different concepts, dynamic range compression and data compression.
The first is like an automated volume control that lowers the volume really really fast in a matter of ms when the volume reachest a certain threshold, (can also work the other way where in increases below a certain threshold, or both).
The reduces the file size, sacrificing quantity if a lossy codec is uses. Lossless codecs like flac are a bit to bit perfect of the original.
It’s not MP3, it’s Ogg vorbis
They’re encoding the top songs with higher resolution, the next bunch ( it’s still millions of songs) with about half that bitrate and not planning to save the next tranche which is like a looooot of songs but each has less than 1000 plays over the history of Spotify.
This is what I vaguely remember after reading the whole original post AND after a night drinking.
No nitpicking.
They didn’t encode/reencode anything with popularity>0.
What they are providing is the same as the available source.
They did reencode songs with popularity=0.
Spotify uses I think 192 or 320kbps Vorbis which is quite good and still has small sizes.
Even that’s 300 TB so I kinda understand. If this is all thar survives into the long term that’s ok I suppose
As I remember no one could tell a different in some testing done vs FLAC and 320kbps Vorbis, so I think its plenty for an archive.
Simone reminds me of the class perspective: musicians here behave like atomized small owners, caught in their enterprecarity, who (legitimately) ask for some defense of their property rights, attacked both by hackers and by the big monopolists of platforms and AI. Because from these property rights, in this case IP, comes a rent, and from this rent, independent artists and label owners try to make a living. Again, right or wrong, this is what’s happening.
I remember reading somewhere that independent artists make basically no money from Spotify.
Is that still true, or have creators found a way to claw back value from the platform, and that’s why they’re defending it?
Spotify performs as one of the worst when it comes to creator royalties and appropriate pay.
It’s not THE worst, and these numbers have shifted since 2022. The general sentiment (that Spotify stiffs artists) is still shared by the general community.
Add it to the fact that, quoting from the aa’s blogpost,
≥70% of songs are ones almost no one ever listens to (stream count < 1000).
The vast majority of the artists are to be “stiffed” of like $3 or less per track during the nearest eternity iff every single one of their listeners is to download the archive and only listen to locally saved copies. Which, given spotify tries to do it themselves by recommending (and, likely, creating) ai slop, tells us more about their predatory practices than of anything else
This plot is complete BS, Deezer pays double than spotify.
And one of the highest payers, Qobuz, is missing
I’m always surprised that Deezer is still around. Who on earth is still using it?
Also, fuck iHeartRadio. Not for their royalties, but for buying up tons of local stations and stripping them of anything that made them even remotely worth listening to. Fuckers bought the only station in town that played any amounts of metal and turned it into yet another top 10s station
My friends and me in the family plan switched from Spotify to Tidal, but due to lack of japanese songs + vocaloid, we have hopped on Deezer instead.
I use Deezer. After Spotify started subsidizing Rogan, it was the only service that had a reasonably Spotify-like UI. Apple Music was absolutely horrible, especially for someone who already has a local library.
I liked Deezer until it’s redesign a few years ago. Now the design language is so ugly I could never consider paying for it. Their biggest flaw for me though is that some albums are just randomly not available in lossless and the only alternative is mp3 which I refuse to use in 2025. Some lossy codecs are great so why do Deezer and Qobuz still mess around with mp3?
I used Deezer up until last month, seemed fine to me
This graph is weird. Soundcloud looks like it pays more according to it, but it needs to swap positions with Spotify.
What’s up with soundlcouds position in that chart?
I think the point is more general about profiting from “renting” their music rather than from their labor. The fact that Spotify gives them peanuts make their position even more miserable.
who (legitimately) ask for some defense of their property rights, attacked both by hackers and by the big monopolists of platforms and AI
This does not sound like they’re defending Spotify.
