320kbps (and Other Lossy Audio Voodoo)
- Posted by mariteaux on September 21st, 2020 filed in Strong Opinions
- Comment now »
I handle MP3s on a pretty regular basis. I have my own vaguely picky specifications for how I want my music collection, neatly sorted in lossless for later use in fucking around (or in the original format and bitrate if lossy to begin with), but I keep a set of casual listening MP3s on this computer and the eMachines box and they do well enough.
In all my handling, I’ve also come across some serious sophistry. I think anyone into this hobby has. Consumer audio is basically the premiere avenue for snake oil today, thanks to the fact that our ears are shit and easily biased by perception. You throw enough money at this hobby, enough gold-plated cables and vacuum tubes, and you too can become an armchair engineer, casually able to pick out the slight distortion and sibilance in the high frequencies with the best of them.
Naturally, there’s a lot of misconceptions people have about audio, especially lossy audio (you know, MP3, AAC, Opus, Vorbis), and they irritated me enough to rant a bit about them. Keep in mind, I didn’t go to school for this, I’m just an asshole on the internet, so for all I know, we all really do need to listen to our MP3s through high-bias 24-bit fiber-optic supergrounded kosher DACs. Somehow, I doubt it.
1. “I can tell the difference between these two high bitrate formats with my $25,000 audio setup”
This is something I see said a lot on various audio forums–basically, some lifelong audio geek has a completely overkill setup and can tell the difference between a compressed bit of audio and an uncompressed one, therefore he’s really cool and MP3s are stinky.
I think the internet has totally forgotten the role that lossy audio was meant to play. It’s lossy; by its nature, using a lossy audio codec means it will sound worse. Whether or not you can tell the difference is irrelevant. The reason you do this is to make the file small enough that people can handle them freely on bad computers and slow networks. These are not audiophile formats and they were never meant to be.
When MP3s first appeared in the mid-90s, the CD (itself considered a perversion of good quality audio for a time) was the gold standard of audio reproduction. There is no compression involved with Red Book CD audio. For all intents and purposes, it is as good as you’d ever conceivably want. (Some people have tried to force the “HD audio” meme which involves higher sampling rates that we can’t hear and a higher bit depth that we also can’t hear; needless to say, DVD-A and SACD never took off for a good reason.) You weren’t meant to listen to MP3s all day; you were meant to use them when you couldn’t listen to your CDs.
In any event, whether or not they can genuinely hear a difference is kind of irrelevant, because you’re not meant to listen to MP3s on the kinds of setups that could conceivably poke holes in the format. They’re meant for broadcasting over the radio, or over the internet, or for use in little portable players where you’re probably too busy listening to train and car noises to care much about some wonkiness in the high frequencies. If you have a good setup, good! Keep going with lossless. That’s what they’re meant for.
Of course, I kinda doubt these people really can tell a difference anyway. To date, no one has provided ABX results that show they can reliably tell the difference between, say, -V0
MP3 and an uncompressed original. With increases in encoder tuning and efficiency, Lame 3.100 actually does fairly well (as in, not annoying, not as in it’s transparent) at 128kbps, let alone higher up. A lot of these threads are dated 2002, 2003, and I’m sure with a modern encoder and the right format (like AAC or Opus), these guys wouldn’t have a clue.
Of course, these people would then say something to the effect of “well, ABXing is useless anyway because it’s like 15 seconds of each, I’ve been listening to these songs for years and I know how they’re supposed to sound.” Yes, real post on one of these forums! And at that point, kids, you just give up and start drinking.
2. “If I convert it to a higher bitrate MP3, it’ll sound better”
To repeat from the last section: lossy is lossy. You will never get the original data back unless you start from the lossless file again.
Converting between compression formats is called transcoding, and lossy-to-lossy (or lossy-to-lossless-to-lossy, I’ll come back to that) is a ridiculously bad idea. The algorithms involved are meant to work on uncompressed audio. Of course, they’ll work on any audio, but the way these things throw out information means it’s infinitely more likely to make the existing compression artifacts even worse. Doesn’t matter the bitrate, doesn’t matter the format (except AAC-to-AAC, for some reason, those tend to be okay on transcode); it’s a bad idea.
I’ve seen one too many very bad sounding MP3s that are, ostensibly, 320kbps, and I know someone got cheeky somewhere down the road, took what’s probably 128kbps, and converted it to 320kbps. This happens a lot with stuff you download from file-sharing networks or strange Blogspots. Believe it or not, even going up to the format’s maximum bitrate, you’ll still make the file slightly worse sounding (not to mention much bigger for no improvement, thanks) by re-encoding. The damage is just likely to be more slight thanks to the higher bitrate.
I know there are fair reasons to transcode, like if you wanted to put a Vorbis file on an iPod, but trying to get better sound quality from an MP3 by re-encoding the MP3 is not one of them. Seek out a lossless or uncompressed copy and use the format and bitrate that you want to use from the start.
3. “If I convert it to FLAC, it’ll sound better”
So you might be even cheekier and think “well, what if I just convert this MP3 to a lossless format? That’ll get the loss back!”
Nope. It’s still the MP3, just now incredibly bloated with no improvement in sound quality whatsoever. It just won’t lose any additional sound quality. The original MP3 will always be the best it’ll sound, and if you want to convert it back to a lossy format, you’ll be transcoding lossy-to-lossy in essence.
This one especially gets to me because of how much larger FLACs are compared to MP3s. If I want to get them small again, I have to transcode–and lose even more quality in the process. Because Bandcamp requires a lossless upload, I’ll see people occasionally do this with their MP3s in order to get them on the service. If you do this–please just include them in the download or offer them zipped separately. I’ll still happily buy the MP3s, I just don’t want to make them sound even messier.
The only time when storing lossy audio in lossless is a good idea is if you’re doing any kind of editing on it; storing it in lossless means you won’t lose any additional quality when you go back to edit it.
4. “256kbps AAC is equivalent to 320kbps MP3” or suchlike
This is another bothersome bugaboo I see posted about sometimes, sometimes by official site documentation even, albeit a more subtle one: the concept that lossy formats are comparable enough that you can equate them at specific bitrates.
AAC is different than MP3. Both are MPEG standards, sure, but AAC has an additional 10-15 years of research behind it. AAC-LC is a leader down around 128kbps, and I’m convinced it’s transparent at 192kbps, based on my ABXing and ABXing I’ve had friends do. Both work completely differently, and when they artifact, it sounds completely different. If you’re using a non-competitive MP3 encoder (like Xing) and a leading AAC encoder (like Apple AAC), the chasm widens even greater.
It might sound pedantic, but it contributes this idea that lossy audio is this simple, that one is equal to this one with these settings. They’re better for different things and format should be chosen based on your needs. Don’t care? I agree! Go with MP3 for everything unless you use Apple products, then go with AAC. Otherwise, the best you’ll get is that AAC is transparent at a lower bitrate than MP3. Entirely different formats. It’d be like saying a q50 WebP is equivalent to a q90 JPEG or something, you just wouldn’t say that.
For the record, 256kbps AAC is just as overkill as 320kbps, and I highly doubt anyone can really, reliably hear the difference between either and an uncompressed original. And speaking of that…
5. “320kbps is/sounds the best”
This shit right here is why I wrote this post. I am so tired of seeing 320kbps MP3s. They’re huge for no gain in quality, you can’t transcode them without fucking them up, and everything, from major labels packing in download codes with vinyl to netlabels offering free downloads, give you one choice in 320kbps MP3s.
Now, why the significance on 320kbps? You see, 320kbps is the reasonable maximum bitrate you can have in an MP3. With a freeformat MP3, you can go much higher, but no player will actually play those. 320kbps is about your limit. Therefore, everyone’s just said “okay, just encode everything at 320kbps! It’ll sound the best, practically like the CD, and it’ll work with everything.”
No.
As I said, 320kbps are a ridiculous waste of space for the fact that they offer no benefit to sound quality over any of the higher bitrates, especially in VBR. At that point, you’re only getting about 20-25% space reduction over a FLAC file of the same song, which loses no quality and can be transcoded. Yet, for some ungodly reason, I don’t get a FLAC download when I buy an album on vinyl. I get a 320kbps MP3 download that both sounds no better than -V0
or -V1
or even 256kbps and takes up more space on my device.
Now, I have had a friend who could (seemingly) reliably tell a 192kbps MP3 apart from the uncompressed original through ABX, but that was in CBR. In the equivalent VBR mode (-V2
), I doubt it’d be the same story. Certainly it wasn’t the same story with a 192kbps AAC file, which she got about half of the trials right. 50% is no better than guessing. Therefore, I know it’s basically transparent at that bitrate, and my ears can’t tell the 192kbps MP3 apart from the uncompressed, so I’m happy settling for that.
Again, this is the point of lossy audio: you find the lowest bitrate, and therefore the smallest files, that doesn’t offend your ears. If you’re looking to “maximize” quality, use a lossless format. Lossless formats are mathematically identical to the CD, though much smaller, and give you the option of converting to whatever bitrate and codec you need, with results as good as that bitrate and codec will ever give you.
If you’re bothered that streaming audio on SoundCloud sounds like shit, go stream audio on your own fucking site, or better yet, sell CDs again! Not like they weren’t uncompressed from the start or anything…