It is not a secret that I prefer lossy. I've read so much into lossy compression when I received the first WAVs and Retromatic's CD rips and what to do with them and how to proceed to build my own music library.
And after a long read I concluded that FLAC is really not worth it.I still am reading into these topics once in a while and nothing has changed in my sentiment.
► I am using MP3 for my audio which in this post will stand for any lossy encoding and is also the lowest denominator, thus the "issues" of MP3 is even less with Opus/AAC/YouNameIt.First of all, when you enter ChunYinZi's question, you will find dozens upon dozens of similar threads, which usually goes similar to this:
● "I've invested money for HQ Audio on [Streaming Platform], and I can't hear any difference!"
● "I can't hear a difference from lossy and lossless"
● "I failed ABX listening test at 192 kBit/s MP3 vs FLAC. WTF!"And these are from HiFi forums, "audiophile forums" and from users quoting their several-thousand Dollar expensive equipment, monitors [professional speakers] and wired headphones with external DAC/soundcards. And after a while there forms a pretty unite picture of the MP3 format - and that it's actually pretty good despite its few flaws. Most use FLAC for "peace of mind" and only a small handful of people actually and truly hear the difference or being bothered about it.
1) "Why would anyone use an old, lossy format like MP3 in 2023?! That format should not be a thing anymore at all!!"MP3s are probably the most wide-spread audio format (except for PCM maybe) and they will still be relevant for the next decades. Just because streaming services or lossless options are a thing does not mean people trash all of their MP3s. Disks unobtainable anymore, rare LP recordings (*looks left and right here on LMT*), personal voice recordings, the favourite podcast from 2007 or the simple fact that MP3 - contrary to (wrong) belief - is a
perfectly fine HiFi audio format will keep MP3s around.
● But why lossy, "disk space is so cheap"?? Well, what's worth the new obtained disk space if you fill it up with barely audible or even outright ultrasonic content? I estimate that it would
more than double my current collection in file size for inaudible audio information. And no, "for the peace of mind" is not a reason for rational-thinking people.
● So, why don't go with a newer format, if lossy, why this ancient format?? The CD standard is over 40 years old, people still listen to LPs - in its core a 136 years old format! LAME, the most famous MP3 encoder, has had its last update
late 2017! This did not rule out the fundamental design flaws out (that is what its successor AAC is for) but it shows LAME has been changed a lot in the past with the help of worldwide developers.
Today, LAME is considered the best MP3 encoder at mid-high bitrates and at VBR, mostly thanks to the dedicated work of its developers and the open source licensing model that allowed the project to tap into engineering resources from all around the world. Both quality and speed improvements are still happening, probably making LAME the only MP3 encoder still being actively developed. ~
Source: LAME webpage, 2017 [side note: They are currently working on a 3.101 release with more quality and speed optimization under the CHANGELOG tab!]
Also,
MP3 is hyper-compatible (see later down this post), and the differences to the competition in terms of files size and audio quality can be neglected.
2) The bad reputation of the MP3 format that does not go away...MP3s are not MP3s. You can have very well-encoded MP3 files and then the ones that make your ears bleed. But it isn't on the format itself!
First, let's start with the encoder and decoder oddity. The MP3 decoders need to follow specifications set very closely. Meaning that playback devices have no real freedom during development. MP3 encoders however have much more leeway. So when you code an MP3 encoder, you can create the worst rubbish in sound and it gets the MP3 stamp on it. As long it plays back, it "passes" as MP3. While the original Fraunhofer codec (FhG) directly from the developers/inventors offer a very high quality and LAME is known to be even more refined - both produce very high-quality results. The worst MP3 encoders are early ones or poorly ones like "Xing" or "Blade". To avoid the high licensing cost, these poor encoders were used in the late 1990s to produce small files for file sharing.
On top of that, generation copies (a MP3 of a MP3), wrong settings (especially amongst the Joint Stereo-haters) and partially bugged encoders caused many poor results. But these were, except for the buggy encoders,
user errors. It's not the steak's fault if the cook has no skill.
Nowadays, you have people claiming MP3 is "lifeless", "cassetty-sounding", giving them "headaches". Just to admit they haven't tried the format since two decades due to the bad experiences back then. The other side claims that only losless audio makes you "experience the music different", that they suddenly "feel the music" more (just to not notice the FLAC source is from a MP3 master, as some labels and distributors even admitted!). The surprise of failing blind listening tests - the realization the MP3 sounds almost completely indistinguishable from lossless sounds - is often big when they actually try out on a modern MP3 encoder. And that results in Reddit threads or this very thread here!
Another thing that I often read between the line is the strange perception that MP3 is the dirty one amongst the virgin (untouched PCM WAV files) and innocent (FLAC) because it's been modified by an algorithm and thus
must be inferior.
3) What the MP3 encoder does to your audioThe goal of this format was to drastically reduce the file size needed while still retaining "good-sounding" audio. Back then, to save precious space, many different methods were used, often reducing the audio quality drastically. One, for example, is the reduce of the sample rate. The famous shooter "DooM" from 1993 for example has some of their game sound effects in 11,025 kHz to achieve that. That is near telephone quality!! There must be a better way! And here the Fraunhofer Institute joins in. They found out what the human ear and human brain can conceive and process -
and more importantly what it can not. The results, the natural flaws of our hearing and thinking resulted in something called a "psycho-acoustic model", or short "psy model". Masked frequencies and, to save extra bitrate, cuts in the higher frequencies that an adult ear can't hear anyways are a fantastic and very sneaky yet ingenious ways to shrink audio while we "hear no difference".
The big, uncompressed digital audio standard at that time, the CD, is a PCM stream of 16-bit and usually stereo. But this contains so much unnecessary audio information and thus bloating bandwidth somewhat.
[I read somewhere to cover the full human hearing with even a bit of a safety buffer would require about 13-bit. So 16-bit is already overkill...] The psy-model of the MP3 encoder takes the source audio (PCM), applies its psy-model and writes a completely new file. Just like a comb removing (mostly) unnecessary audio information, the left overs are what are "really needed" to rebuild the audio for our ears and brain, using all the little tricks about our flawed human perception. Depending on the settings or the codec used, this can be very severe (below 128/96 kBit/s) or basically inaudible (V2 or >256 kBit/s).
All the other stuff is removed - and that is the reason why MP3 sounds great with such small bit rates. Generally spoken,
LAME-encoded MP3s are "transparent" at 200 kBit/s, and VBR1, VBR0 (that is what I use) and CBR320 kBit/s are actually not recommended because it only increase file size with low gain in actual quality. Or in other words:
At 256 kBit/s a proper MP3-encoded audio file is indistinguishable from its original, uncompressed source (minus very specific "killer samples"). And that is a nice feat in my opinion!4) How a MP3 file still sound so goodI already mentioned the trickery that a MP3 file really is - the remains of an audio stream that we can actually hear and process while scrapping the rest with a hint of "the brain will fill in the rest from experience". For example the frequencies being cut. In lower bit rate settings, there is a hard 16kHZ cut-off. But have you ever looked at your stereo's spectrum analyzer where the vast majority if your music is being located in the audible frequency range? Only a very few instruments reach such high frequencies at all (usually cymbals or hi-hats), and they are often masked, and after all that, the average adult human won't hear that high frequencies at all anymore.
So why bother if these frequencies are gone? If you allow the encoder to spend more bits during encoding, the low-pass filtering is not even a thing. A 256 kBit/s MP3 usually has all the frequencies intact - at least as much the MP3 format can store it (the biggest flaw of MP3 is the inefficient storage of the very high frequencies). But the biggest "body" of a music track is in the lower and middle frequencies. And there, MP3 shines:
With just a mere 192 kBit/s, it can build a surprisingly faithful audio stream in stereo.
I mentioned masked frequencies - this sound so off-putting. But imagine you have a radio turned on and then switch on your vacuum cleaner at the same time. The radio, while actually still playing back music, will be masked by the vacuum cleaner. There you go, "masking". The masking is again a cool trick that uses our brain's poor judgement and great ability to fill in gaps that are not actually there. For example in the track "Line Feed" from the SONOTON album "Sound Force" there is a great example. The song starts with massive synthesizer "clicking"/trickling. High frequencies all over the place, and despite my encoder is set to V0 which is usually ranging from 240-280 kBit/s, it is constantly at 320 kBit/s. Until the synthesizer "flute" kicks in. This synthesizer flute masks the trickling synthesizers. Because these are now less audible, the MP3 encoder can assign less kBit/s to these portions.
However, our human brain does not realize any difference! For us, the "clicking" keeps going at the same volume.As soon as the "Synthesiter" fades out, the "clicking" is more audible again and the encoder asks for more bit rate again.
► For us, there is no/barely any audible difference - especially not if you don't know what has been cut in the first place.
► For the MP3 encoder that constantly seeks for things to sneakily cut from the audio to save bandwidth, it will cut there without us noticing.
Smaller memory/bandwidth savers are for example "Joint Stereo". The audio quality is the same than "full stereo" (L/R in strict channels), but due to mathematical/logical shenanigans, you can either save bit rate or assign those bits more to other portions of where it's more needed.
So, we have our ears and our brain being pretty successfully fooled by the psycho-acoustic model of the MP3 encoders. But what about your playback hardware? No, I don't mean the ridiculous exaggeration of the likes of "Of course, in the bus with your Bluetooth earbuds, you can't hear a difference". I have read about users of expensive headphones and proper hardware "chain" admitting they either can't hear any difference or it's so subtle and takes lots of continuous listening effort to make out.
5) Problematic cases with MP3These are very rare. Well, let's say:
In reality, these are rare. I have seen users from other forums deliberately throw "killer samples", often even test tones that never happen like that in nature at the MP3 encoder just to point fingers. I can't remember when I was listening to a static square wave test tone... Anyways, in my entire V0-encoded library I have exactly one track that has issues. It is a slowly building up reverse cymbal. At about halfway through this build-up, it starts to sound a bit "hissy" or "jagged". The WAV which I had and alternated between and the MP3 in Audacity had a smooth crash cymbal. Emphasis on "a bit". The problem is there, but only because I knew from the direct comparison with the WAV file. "Naturally" happening crash cymbals or constant hi-hats sound surprisingly well considered these are the weak points of the MP3 format.
Apart form low-bit rate artifacts, there were other problematic samples like "pre-echo" and such things. While I listened to these samples, these other problems are either fixed by using a reasonable bit rate (over 200 kBit/s) or by a change in the encoder or its setting. Modern MP3 encoders still have flaws that are just part of the format's design, but they have been either reduced heavily in occurrence (how often/when these happen) or by intensity (how strong these happen).
But, in my experience with the format, in the very, very vast majority of the cases, you will never actually run into such a problem sample, and even if, it is by far not as bad as people claim.
6) But what if you want to convert? From FLAC, you can encode easily without further loss! That is future-proofing!This is something I never really understood, and it comes up in almost every thread. The question in the first place:
Why would I want to convert my MP3s anyways? It is a perfectly fine end-user audio format. And due to its widespread use and the lack of any royalties to pay, it will be supported for a long time.
Also, I found that I have to convert other (lossy) formats while my MP3 does not need any further encoding. For example:
● Dropping an AAC into Audacity - "No ffmpeg library found. Can't import"."
● Dropping an AAC into Magix Video Deluxe - "Sorry, purchase that MPEG codec pack for 3.99€".I ended up transcoding these AACs, OGGs and OPUS files - all a non-issue with MP3 which is accepted by all and everything. If your toaster would have speakers, it would accept and play MP3 files before anything else. And these above are computer programs. How about your car stereo or kitchen radio?
MP3 is hyper compatible, and thus this whole "what if you want to convert MP3s" is a non-issue. And even if I would have the need to change formats, I can always create a PCM file out of my V0 transparent MP3s.
7) Dude, this trickery of MP3 is disgusting. Give me the real audio. FLAC is the only real way of archiving......says the person using vinyl records or other analogue formats that are comparably very inaccurate but somehow enjoys the "warm sound" [read as: Changed audio compared to the studio master] of it. The same person complaining about subtle, barely noticeable changes in audio of a lossy encoding.
Let's be real for a moment:
FLAC is a "sterile", bit-perfect file format that should be used with other bit-perfect formats or media, like CDs. Not tapes, not records. These media formats are far from the level of perfection of FLAC/WAV could provide. People FLAC'ing their 40-year old records -
maybe even in 24-bit while using a jack-to-jack audio cable that introduces a subtle hum- often don't know what they are doing. I am not sure if I want to be bothered by the opinion of people not really understanding the advantages and ingenious ways of how lossy encoding works but then categorically deny any form of lossy compression anyways and insist of using overkill-FLACs..
On LMT in particular, MP3s are not necessarily hated, but people would prefer the FLAC from a damaged/worn-out analogue medium with surface noise over an objectively better-sounding MP3 just because it is a lossy encoding. I have Parry Music MP3s from the BMG website (256 kBit/s) that sound superior to the LP recordings by "Del Frost", but those inferior recordings are in FLAC, so I never bothered offering these MP3s because I know they are frowned upon . . . It is basically a different form of upscaling - creating 24-bit FLACs out of an analogue, worn-out recording.
Ironically, digital sources like the CD audio are working very well with MP3 encoders, taking away my incentive to use FLAC even further....