Thoughts On FLAC

Thoughts On FLAC

I’m a bit of a curmudgeon (ok, a huge grumpy bastard) when it comes to the audio format FLAC.

Here are some reasons why:

Objective

– FLAC offers, at very best, an extremely marginal quality benefit over 320kbps MP3.

I’m being charitable here, as I personally don’t believe that the overwhelming majority of people (i.e. basically people who are not audio engineers on a good day) would be able to tell the difference in an ABX test. My suspicion is that some mastering engineers listening loud on high-end monitoring setups would be able to do it some of the time.

For real world situations, I believe 320kbps MP3 paired with a decent headphones for PC gamers is sufficient.

– FLAC is more work for me!

It is just simply more work to support FLAC every time I release something. I have to encode again, listen through every track again, tag, archive and upload again. I have to create separate products in Fastspring etc. etc. Obviously this isn’t THAT difficult, but it is objectively more work.

Subjective

– I really hate tagging audio files: I actually find that process immensely tedious and slightly depressing. I know what a huge over-reaction that is, but I’m being honest about my subjective responses here. I haven’t got a system which enables me to tag an album once and then just copy all the tags over to a FLAC version – if someone knows of some software which is good for that, it would help me out.

– I didn’t think FLAC was genuinely lossless, but I was wrong about this so perhaps I should change my opinion!

– I suspect FLAC to be the preserve of what I call “golden audio cables people” i.e. people who don’t understand that the 80:20 rule applies to audio setups, display gigantic amounts of confirmation bias when it comes to various technologies and spend their time talking about weird, pendulum-wavy stuff that real audio engineers find incomprehensible. They often don’t understand that even a 1% difference in how an album is mixed would have 10,000x more impact then any of their fiddling. A lot of them listen to music on hi-fi’s rather than studio monitors, and then complain about accuracy.

This is a stereotype, of course, and grossly unfair to a lot of people. As soon as I hear FLAC, though, this part of me activates.

– The only comment I received from some fans was annoyance that I didn’t release something in FLAC.

The overwhelming majority of people who like my music seem unbelievably nice. I have received so many positive tweets, emails, comments etc. that I have really changed the way I feel about my music and gained a huge amount of creative confidence in the last few years. These comments inspired me to go on with projects like _ensnare_ and to push the boat out with a lot of the Frozen Synapse music.

So the contrast when you get “WOT NO FLAC???” or worse posted on something you’ve just released is such a huge let down that it has definitely predisposed me to hating FLAC. Even a comment like, “hey, I like this album, but I’d really appreciate a FLAC version if you had time. FLAC is good for these reasons: x,y,z” would be a VAST improvement. I know that this point is shot through with the “I wish people would be polite on the internet” fallacy, but there you go!

I’m going to continue supporting FLAC, especially now that I know it’s genuinely lossless, but I’d be really interested to see any actual evidence of scientifically conducted ABX tests.

When a difference in audio is the subject of genuine scientific uncertainty, I would be willing to bet that normal people in normal environments cannot discern it!

However I could well be wrong – let me know what you think. “I tested this myself and I could hear a difference” is not going to convince me, sadly, because I don’t know how good your test was.

Finally I want to reiterate that FLAC is clearly BETTER than 320kbps MP3. Its superiority as an *archival* format is entirely clear, and that argument makes more sense to me than “it sounds better”. Everything should support FLAC – if a superior format exists then I shouldn’t need to be using both in this day and age: that’s what irks me!

41 Responses to “Thoughts On FLAC”

  1. Brownd:

    I’ll just leave this here…
    http://productionadvice.co.uk/why-mp3-sounds-bad/

    Nevermind the title, this guy is a professional mastering engineer, founder of the Dynamic Range Day.

    To make it short:
    · By definition it loses information on the music, mostly small details
    · Might add distortion to the track

    These things depend a lot on the encoding process and bitrate, though. They’re not as noticeable at 320kbps as in 128kbps, for example, but that doesn’t mean they’re entirely erasable.

    I think the issue some people seem to have might be the fact of purchasing music with limited quality. When buying music many people actually do it to listen to it with total clarity as it should be (e.g. CDs, the lossless digital audio standard), so it’s kind of a letdown when digital purchases often come in lossy formats like MP3 only. They would never be able to listen to it in better quality if they feel like doing so. That’s why supporting FLAC is a great thing.

    And about the tagging, what program are you exactly using for that? Because there’s programs that let you simply copy-paste tags between sets of files (e.g. copying an album from a format to another)

  2. Brownd:

    There’s really not much science needed about why FLAC is better, because its principle is simple: it’s a zipped WAV file. That means you can unzip it without loss, and even burn it to an audio CD.

  3. Paul:

    I’m not interested in the science of why FLAC is better: I know it’s better. I’m interested in the science of whether anyone can tell the difference or not.

  4. daggity:

    Person who bought the FS OST just for FLAC (After having it in MP3 from the HiBFS). Doing some (obviously not very scientific) AB tests on the internet in regards to quality differences, I could hear them. I definitely heard a difference between the FLAC and MP3 versions of the FS OST as well.

  5. Paul:

    Brownd,

    If you just left it here, then I’m a bit worried about reading it. What if you left it here by mistake?

    Despite invalidating your entire comment with awful internet banality, you linked to something interesting. Unfortunately, that post is horribly fallacious: he makes all of his arguments about 256 kbps MP3 then suddenly posts a flustered edit about how “320 is SOMETIMES ok”. It’s entirely unconvincing. He should be making those arguments with WAV vs. 320kbps MP3 throughout.

    Nobody reasonable could argue that anything less than 320kbps, encoded properly, is acceptable. I’m not arguing that.

    Your point about wanting to buy something without lossy compression for archival / future conversion is definitely the sticking point for me: that’s the good reason for wanting FLAC in my opinion.

    I’m using MP3tag – it sucks and I hate it. What would you recommend?

    Also DYNAMIC RANGE DAY? Come on!

  6. Paul:

    Daggity,

    Could you post an audio example of where you could hear differences between the MP3 and FLAC? I’m genuinely interested, firstly because I find that a bit dubious, and secondly because I could be encoding incorrectly in some way so I’d like to check it out.

  7. Benji:

    The only time I have a problem with offering FLAC is when on caving, authors are re-encoding a lossy format (usually 320kbps MP3) into FLAC.
    I’ve seen it happen, and then argued the point against people blind to it because ‘it’s a FLAC file therefore it’s better’.

  8. NickD:

    The argument of fidelity carries through to all forms of media and has done for a long while. I believe that it is the job of experts to be advocates of quality as it helps drive technology forward.

    I agree that the majority of people on the majority of playback devices would not be able to hear the difference. But this isn’t particular to audio. For instance, I would struggle to be able to tell the difference between two SLR lenses because I am not a photographer, but I could hardly argue against the legitimate need for high-quality lenses. Similarly, someone new to audio production probably wont be able to tell the difference between a cheap compression plug-in and a vintage Fairchild. In fact, they might not even be able tell when dynamic compression is happening at all.

    Of course, there is plenty of audiophile nonsense out there. Indeed, there’s healthy money to be made by snake oil salesmen. However, this does not invalidate the core necessity for us to give a damn about quality.

    If you’ve gone to the (considerable) effort of writing a piece of music and you’ve taken care to get the best production possible, then there’s no reason in the world not to want there to be a version of it available that sounds as good as it can. Why not aim for perfection?

  9. Lionel Hartman:

    I often listen to music with my dog and he appreciates the high dynamic range in frequency that the FLAC format provides.

    There is a large market for dog music and I find it offensive that you have deliberately overlooked it in your arguments.

  10. Paul:

    @NickD what about when I haven’t taken care to get the best production possible? People are arguing this with me about a very low-fi album…that’s one of the things I find silly. If you care so much about this then surely you should care about the frequency range I’m using in general, or my use of compression etc.?

    @Lionel Hartman

    I think I love you and your dog.

  11. Paul:

    Obviously taking an MP3 and then encoding that to FLAC is inherently ridiculous.

  12. urbanhusky:

    Hi,

    I feel partially responsible for this so I hope to bring something useful to this debate 🙂

    Concerning your objective view on FLAC: I cannot support your view that it offers only a marginal benefit in terms of quality. Granted, this benefit, regardless how small, is what drives me, what makes me enjoy music even more so there definitely is some subjective weight to it.

    I can understand that FLAC is more cumbersome to deal with, but maybe there are ways to improve how you handle it?
    Why not create the FLAC version first and then convert it to mp3? Most converters should be able to copy the tags for you 🙂
    Also, personally I wouldn’t mind paying extra for FLAC – if the price is right. For me, paying 1.50€ per track (as how it’s done on beatport iirc) is just too much. Raising the price of a 6$ album to 8$ for FLAC for example could work for me 🙂

    While I consider myself somewhat of an audiophile, I am very sceptical about actual gains or changes in quality – for example when it comes to USB or network cables. I have a rather analytical and scientific mind and world view, so when I think I hear a difference, I want to be sure that I do, in fact, hear a difference (ABX tests etc.) and try to figure out the reason as to why this difference occurs.
    I hope to take a look at audio cables etc. in the distant future by investing in an oscilloscope, function generator and such – go at it from an engineering perspective (and even then, seeing some changes do not automatically mean that they are audible).

    I’m getting sidetracked here. Using a Native Instruments Audio 2 DJ, which is a good sounding but by my standards not perfect sound card, I did an ABX test last year between 320 kbps mp3 (lame -q 0) versus FLAC: http://nifelan.de/hermes/get/63c96c51d5628/abx_2011_320mp3vsflac.txt
    ABX test was done in foobar 2000 with the ABX plugin. The audio setup was the Audio 2 DJ via ASIO driving a Sennheiser HD 25 II.
    While this isn’t a perfect result, I tend to think that it is significant enough to show that I, personally, am able to hear a difference between 320 kbps mp3 and FLAC.

    I do have a small problem with ABX tests though. I’m not questioning their validity, however these tests are _hard_. I usually notice changes in music quality after I’ve listened to the same album for weeks at a time and discovered new things and aspects about it. If the same things don’t show up in the previously used setup (again, ideally ABX), then there must be a difference. However, noticing these changes in the very short time frame available for ABX, can be very difficult – which actually does speak for the quality that is achievable with mp3. Yet, this difference exists and can be heard – and with time appreciated even more. As far as I know, ears can be trained to listen for such differences and details.
    Another reason why I dislike mp3 (and other lossy formats, less so ogg for personal preferences though) is the introduction of artefacts.

    When listening to music, I get the feeling and imagination of all different kinds of shapes. Taking my ABX test for example: the first few seconds of the track (up to 0:27) were the most telling part. The guitar strokes appeared very sharp with various small spikes on top with FLAC – while in mp3 it appeared more like a very round and not as defined high top.
    These fine details, which – to me – are lost with mp3 are what drove me to the decision to collect music only in lossless formats.

    This brings me to the point of the behaviour of us FLAC loving people. I think part of the anger and hostility some people exhibit could be because of the following:
    Imagine, finding a great artist, an album that could sound great, bring joy and take your mind on a thrilling roller-coaster ride, prompting you to go again and again – at least until the neighbours start to complain and you consider drowning out their disapproving noises by increasing the volume even more – only to find that the album in question is only available in mp3, which feels like being declined a chance to listen to the album at all.

    Granted, this is no excuse for impolite behaviour. I myself may have acted up a bit because of this and I want to apologize for that.

    Another thing that I’m equally as enthusiastic about is good mastering – basically anything but what is available commercially today I fear. Lots of clipping, heavy compression that makes me cringe when suddenly all details are drowned out by noise… I’m not a fan of the loudness war.

    regards,
    urbanhusky

  13. daggity:

    There are two differences I find between MP3 and FLAC in general. I’ve been trying to think of words to accurately describe the differences, because they’re obviously rather minute. There’s a certain fullness that FLAC tracks have, as well as being a bit more clearly defined.

    I don’t think there’s anything in particular wrong with the MP3 tracks. I had listened to them for a while before opting for the FLAC download. I doubt I’d even be able to hear differences on my travel headphones, just my main ones (Sennheiser HD 555).

  14. Paul:

    Hey – apology accepted etc.! I was just interested in exploring this a bit and being grumpy about something in public is sometimes better than just doing it in private as you find out intriguing info.

    I can tell you right now that test is not statistically sufficient to prove that you can reliably tell the difference. That’s why I’m not really interested in seeing people’s personal tests – I’d really like to see any formal research into this. I want a study or review that I can send to someone who knows about science and have them tell me how sane it is. If it’s demonstrable that people in general can tell the difference (even in an optimal listening environment only) then that would change my opinion.

    I’d doubt that I’d be able to tell much about a recording on those headphones as well!

    I think what you’re doing is attaching an emotional value about the importance of sound quality to FLAC and that this is colouring your judgement. I’m not being patronising – I’m sure I use plugins which do the same for me (i.e. they offer virtually no discernible difference but make me feel better about the sound) – I honestly think that’s what’s happening here, and what is happening with a lot of FLAC fans.

    Good mixing and mastering are really important – as you rightly point out. I think I would have had more time for a comment like, “I’d like to hear this type of music but with better production”, which is what I went for in my later _ensnare_ albums than “this should be in FLAC because I want to hear the high frequencies better”. It’s kind of supporting the argument that the difference between MP3 and FLAC is serving as a more of an emotional hook than a practical one.

  15. urbanhusky:

    Regarding the loudness war/good mastering thing:

    For example, take these two recordings of the same song – one released 1991, one 2003.
    http://nifelan.de/hermes/get/9ad7212b23f02/queensryche_operation_livecrime_remastered_vs_old.jpg
    All the details are gone. It sounds even worse than it looks (at the same effective volume).

    One of my favourite “bands”, Infected Mushroom, are also guilty of increasing the volume/compression with each album, going as far as having severe clipping. While this might be a stylistic choice, I do not approve of it because all the details of the various “instruments” are lost and drowned out in the general noise/volume of the track.

    I want to listen to it, be able to turn it up, be engulfed in music – but as soon as I do, it sounds terrible… a prime example of this is Chemical Brothers – The Boxer. Starting 0:39 the track is an incoherent mess.

  16. NickD:

    @paul:

    That’s a good question. I think ultimately it’s a matter of artistic intent. Producers often exploit different formats in an attempt to capture their particular properties, for example bouncing a drum buss to tape and back-again. At this point the argument starts to get dangerously near phenomology which is its own rabbit hole, but linked. Now, if you’re putting tunes out in 320MP3 as part of an aesthetic you’re aiming for, then it’s valid by all means. Hell, there’s still people who put stuff out on cassette tape! Otherwise I would say that any data compression acts contrary to the original intent.

    I would agree that certain programme material is more (or less) susceptible to data compression. I suspect your album would be less affected by compression artefacts than say a perfectly captured orchestral recording.

  17. Brownd:

    Sorry about my opening sentence. I really should stop doing that.
    I use MP3Tag as well. It’s extremely reliable for me, I don’t know what could be your problem.
    When you’re done tagging the MP3s, you can select everything and then “copy tags”, then open the FLACs in the program, select them and “paste tags”. Just be sure not to close the program in between or you’ll have to copy again.

  18. urbanhusky:

    The thing about doing a study with a lot of people is: when 99% of all people cannot hear a difference, this doesn’t mean that I personally can’t. This is possibly also why a lot of us “audiophiles” are so defensive about this. We kind of don’t want to have others define what we are.

    I’ll try to repeat the test with when I have my better soundcard back.

  19. Paul:

    Hey Brownd! Thanks for that – I believe I tried that before but will give it another crack. I really need to find something better than MP3Tag – it really is a crock of shit.

    NickD, I would actually say that 320kbps was part of the aesthetic of No Carrier Rush i.e. the quickest easiest way I could get something out there! I should start using more artistic justifications in my FLAC-related laziness!

    @Urbanhusky

    Sure – I am interested in the consensus. But also, you *personally* would probably need to do hundreds of tests on this to reach a level of statistical certainty.

    I should probably do some listening tests myself actually.

  20. Paul:

    Also with regard to loudness, I totally agree that certain material needs to not be limited to death, even purely electronic / dance stuff.

    There is some stuff on the radio at the moment that I literally cannot believe: the Nero album for example was completely eye-popping!

    It is a stylistic choice for some artists. I can’t imagine a Skrillex track that isn’t ear-bleedingly flat, but he’s definitely “using his mastering chain as an instrument” – it’s an integral part of that ultra-harsh, digital sound and thus completely fine. His track for the Syndicate trailer is, I think, the most clipped commercial recording I’ve ever seen.

    With bands, I’m in two minds. I’ve heard some rock stuff that’s been mastered “properly” and just has way too much detail when it needs more impact to stand up in a modern context – it actually needs to be slammed a bit. I’m not talking clipping here – I just felt that the limiter could have been pushed a bit more. However, for more detailed intricate stuff, then of course you don’t want full-on brick-ness.

  21. urbanhusky:

    The great thing about scientific tests is, that they are repeatable. The only thing that is debatable, is what could be considered statistically relevant.

    My 2011 ABX test has 10 tests in 9 of which I “guessed” correctly – that’s within 95% confidence level. I’m currently in the process of doing another (yay, repeatable) and am at 9/11, again within 95% confidence level (according to Wikipedia). I sometimes convince myself that I hear a difference at the very first second and eagerly press the according button though 🙂
    I’ll try to continue the test but I’m noticing that this intensive listening is very fatiguing – I’ll post the results later anyway.

    Screenshot of the progress: http://img35.imageshack.us/img35/7811/abxinprogress.png

  22. urbanhusky:

    As promised:
    https://www.dropbox.com/s/sf5monn3ggdwg11/2012_06_ABX_urbanhusky.txt

    I have to go now because I have to be somewhere – and listening is currently a bit difficult because of outside noise. I’m also exhausted and in desperate need of a non-music break.

    I fear I won’t be able to listen to this song again for quite some time. ABX ruins music – because you listen to it again, again, again and again – and again 😉

    This test is within 95% confidence level and the combination of both my ABX tests results in 19/23- also well within 95% confidence level.

    Granted, the differences are sometimes very subtle. In other moments it was very clear – guess it depends on whether I start playback with B or A – it seems easier for me to notice the increase in detail than the lack thereof.

  23. Jeremy Gray:

    I pay a premium for FLAC simply because it has been through no lossy compression and I can choose to batch recompress it into whatever format I need. There was a period in time when my portable music was in mp3, now it is primarily m4a, each phase going through different bitrates and settings over time (is anyone else here old enough to remember when the early encoders were so touchy that you had to try both joint stereo and dual stereo to see which generated the best results for a given piece of music?) No matter how many times formats change, or even how many times my chosen bitrates in a given format change as device storage grows, with FLAC I can target the desired format, bitrate, detailed settings, whatever, and batch convert, resulting in files with only one step of lossy compression no matter how many times I do this.

    On another note: Paul, please bone up on batch tag-preserving audio file conversion tools before you post and on basic statistics before you comment. Some of the commenters are not exactly coming across as informed, unbiased experts but this is your post on your site and, well, you’re not doing any better.

  24. Paul:

    I’ve already said that I think FLAC makes sense for archival purposes.

    I still don’t understand why someone would focus on the FLAC vs. MP3 issue with an album that is demonstrably lo-fi in its nature.

    What do you feel I need to know about batch tag-preserving audio file conversion tools? If I need to know more about them than I currently do, doesn’t that support my assertion that I need to do work?

    What do you feel I need to know about basic statistics? I hardly know anything about statistics; however, I do know that a blind test where someone tests *themselves* on something 10 times is not going to be considered seriously statistically significant or valid in any way other than an interesting piece of anecdotal evidence.

    Don’t you think it’s this kind of self-righteous hyperbole that has put me off from engaging with FLAC in the first place?

  25. Paul:

    Meh, I’m conflating statistical confidence with poor testing methodology – I apologise! Apparently 10 times is enough (http://en.wikipedia.org/wiki/ABX_test#Confidence), which I find MIND BOGGLINGLY ludicrous and need someone more intelligent than me to explain at some point – I will seek this out. I leave my earlier error there for all to witness. You could have pointed that out in a less cunty way though.

    My problem with this kind of personal testing is that it’s subject to an enormous amount of external nonsense: maybe the files you’re listening to aren’t perfectly encoded; maybe you’ve heard them so many times that you are more aware of the extremely subtle differences that wouldn’t be parsed on normal listening; maybe you’ve trained yourself rigorously to tell the difference…that’s why I didn’t want a load of “I HAVE DONE THIS TEST AND I CAN TELL THE DIFFERENCE” responses – they’re inherently meaningless other than being, as I said, anecdotally interesting.

  26. Paul:

    I tried an ABX test with two of my own tracks in Foobar – it said the probability that I was guessing was 30%. I’m probably just worse at this than Urbanhusky!

    Urbanhusky, when you’re doing this, are you listening to the same section of each track (i.e. using the “set start”) thing, or are you playing them through from the start and swapping over?

    Can you tell the difference between MP3 and FLAC absolutely cold: if I just gave you two different songs, one MP3 and one FLAC; do you reckon you could repeat that feat under those conditions? I KNOW that I wouldn’t be able to do that currently – I’d have to be specifically listening for MP3 artefacts if I had any chance of accomplishing it as well.

    I can say, with “confidence”, that a 1db change in high frequency EQ makes more of a difference than this – you should definitely be arguing about that and not this if you’re talking purely about audio fidelity.

  27. vdrandom:

    First of all, you don’t have to fill tags twice, just use the converter that can convert tags with everything else.

    Also, I do prefer to get FLAC for one and only reason: if I buy a CD, I get 44,1kHz 1440kbps media which I can rip the way I want to in a format I need. And when I am to buy an album via Internet, I want to be sure that I can rip it on my own rules, including but not limited to Ogg Vorbis (it is better than MPEG 4 Layer 3 on lower bitrates, but that’s not the main reason – Ogg Vorbis is open source and always uses UTF-8 for its tags – unlike mp3). I hate the idea of transcoding from another lossy format – especially mp3. I know that I might not be able to tell the transcode from the original on my current audio devices (hell, I managed to use HE-AAC 64kbps on one of my previous phones and be fine with it), but redundancy is not always a bad thing you know.

    Especially when I’m charged as much for digital download as for physical medium with higher bitrate. That’s just unfair.

    Ah, one more thing, almost forgot. I once read some audiophile article about how awful mp3 is and how wonderful lossless sound is. I’m a sceptical person, so I decided to give it a try. I encoded some files in high bitrate lossy formats (aac 320 kbps, mp3 320 kbps, Ogg Vorbis 320kbps fixed, Ogg Vorbis q9 and even mpc). Then, I decoded them into wav (including original flac) and put in different directories with same file names (so that I see only track names in my playlist, but can restore the path if I look into track properties). After that randomized playlist and listened to it. So, here is the result: I can’t tell FLAC from mp3 320kbps or Ogg Vorbis q9 on my current audio devices. It’s even funnier that I’ve tried lower bitrates and didn’t hear any difference too (mp3 failed on 192 kbps, aac and vorbis – a bit lower).

    Yes, that’s right. I’ll repeat it again.

    I cannot tell the difference between FLAC and MPEG 4 Layer 3 encoded in 320kbps or Ogg Vorbis q9.

    It doesn’t mean that I’m ok with being ripped off from being able to rip (pun intended) music the way I want to. Especially when I’m paying for it.

    Also, [not that] important thing: the older a man becomes, the thinner his hearing amplitude is. Older people can’t hear higher frequencies, so in ten or more years I wouldn’t be able to tell the difference between even lower bitrates! Isn’t that awesome? 🙂

    Also, mp3 sucks (fixed bitrates, problems with tag encodings, licensing problems and so on).

    PS. Excuse me if my English is poor – my skills are a bit rusty.

    PPS. Thanks a lot for letting us have lossless versions of your tracks with Frozen Synapse OST even if you don’t see reason in it.

  28. Paul:

    I am just waiting for the day when storage and data transfer are good enough so that everyone can just be satisfied with WAV.

  29. vdrandom:

    Not WAV please… It doesn’t store tags after all. 🙂 FLAC does and it is an opensource format.

  30. urbanhusky:

    On FLAC vs mp3 regarding an already “low quality” source:
    I’d say that FLAC would be even better suited for the job, because it retains the most information and doesn’t introduce additional errors. I’m also somewhat of a collector so I want to have my music available to me in the best possible format – with my own experiences regarding mp3 I don’t see it quite as important or worth to archive as FLAC.

    On testing methodology:
    I did my tests with what is, to the best of my knowledge, usually considered the best possible encoding for mp3 – by using LAME with the -q 0 option. I do believe that I can repeat the test, but I have to admit that there might be some days where I’d be hard pressed to notice any difference at all – just like a runner might be able to achieve a world record sprint one day and fail miserably on the other, I do believe that the state of my body and mind can change as well regarding my listening abilities.

    The point about having heard a track so often that even the most subtle differences are easy to pick up is exactly what I mean and what drives my enjoyment for music even further. When I suddenly notice that there is an entire instrument, a few more – and even different notes and more layers than I was aware of before, then this brings thrilling joy and lets me listen to almost an entirely new piece of music with new found love and respect.
    I first started noticing this, when I switched from 320kbps mp3 to 320kbps ogg on my Cowon iAudio U3. I listened to the same albums for months at a time. Suddenly, while on a train, I heard details I never heard before in my music.
    When I switched to FLAC later, the difference was yet again so great, that I stood in the train, frozen mid-step with my jaw wide open in surprise. I later went on to confirm my findings via ABX tests btw.

    I also have to say, that I’ve been pretty much trained to listen for details since childhood. My dad played music with electrostatic loudspeakers, which, I think, are well known for their highly detailed high frequency resolution. I learned to play the violin, played it for many years; My own experiments in the past years regarding improving sound quality (better headphones, better sound card, switching from Winamp to foobar 2000, using only lossless music) continue to train my ears to listen for those fine details.

    I have to admit, that I’m most likely not able to tell the difference between each and every piece of music in my collection. I’d most likely would have to listen to it over the period of weeks or even months before I’d pick up changes. As far as I know the short-term memory regarding hearing isn’t that great.

    I found ABX tests to be “easiest”, when I play a specific portion of a track (<10 second duration) from 3 to 10+ times before switching the sample (mp3/FLAC – or X/Y) and repeating the process. Usually when I'm back at the first sample again (so having listened to the same few seconds for 9-30+ times) I can then tell the difference. Another way is to start with the mp3 and then switch to X, then play the mp3 again, then switch to Y – X or Y will have a huge difference somewhere. In the case of Karma Police, the guitar in the background is much clearer and the metal strings can be heard when struck – something which is less defined or even missing in the mp3.
    After having done one test, I then pause for a few seconds (or minutes, depending on how I'm feeling) and repeat the test with a different section of the track. If I stick to the same section of the track, then everything starts to sound the same and my brain feels mushy 😉

    I'm not claiming to be able to notice the difference between mp3 and FLAC each and every time, with every audio setup and every type of music – but like I said, being able to rediscover my music, find new nuances – that's something that drives my enjoyment to absurd highs. Music also has a very strong emotional impact on me and perhaps some of the details that are lost with mp3 are more of a subconscious, emotional component than actually directly audible things.
    So far I find mp3 to be lacking a lot of the high frequency detail – forgive my wording here but for the lack of a better analogy, I'd call mp3 "slower". When I hear each finger hit metal strings, notice every vibration and the subtle vibration of neighbouring strings – such details are often softened in mp3 – at least it appears to me this way. The bass also seems to suffer a bit, loosing a bit of well defined start and end. It's possibly hard to explain this because a lot of this is also coupled with the abstract imagery that music creates in my head 🙂

    Of course, the actual sound setup used and the mastering of the tracks itself might create a much bigger change than mp3 vs FLAC – but when comparing codecs, both the mastering and audio setup usually are identical for both samples, therefore negating any colouring that the sound setup might do because it is applied to both samples. Granted, it might emphasise differences in certain parts of the spectrum – but the claim that there is no audible difference between mp3 and FLAC then cannot stand in the context of that particular test setup.

    I'd love to hear neutral studio monitors and compare them to my Tannoy mercury m1x speakers. My brother has a pair of AKG k701 which are used as reference monitors. While I love the high resolution of those cans, I miss the powerful low end and (very controlled) punch the Sennheiser HD 25 II offers. So far I haven't found headphones which combine the power of the Sennheiser with the accuracy and resolution of the AKG.

  31. Paul:

    That’s really interesting! I’m actually glad I started this discussion.

    I certainly think you’d beat me in a “listening contest”, if such a thing existed! I found the test to be really difficult: interestingly I actually found that I was better at telling the difference when I listened to the same short portion of the track A/B’d with the other sample.

    I have to say, though, is that I would definitely still characterise the difference as “marginal”. We are in the realms of tiny, tiny changes here.

    As you say, I think you need to have a trained ear (i.e. know what you’re listening for) and be in “peak condition” to hear the differences consistently. I think this really reinforces my point, to be honest.

    I found the easiest thing to listen for was very very slight high-frequency aliasing (as you might expect) on the absolute top end. There was one portion of the track I was using with a white noise sweep in it that seemed to be most distinctive.

    I do wonder if different encoders colour this test – I think there might have been a tiny level difference between the MP3 and FLAC but couldn’t confirm this.

    Definitely learned a lot from exploring this – thanks for the comments everyone. I genuinely didn’t think that 10 repetitions of this kind of test would provide any kind of statistical confidence but I was totally wrong about that, so again I apologise for being presumptuous about that.

  32. Paul:

    This “testing 15-ish times is sufficient” thing…having thought about it more, surely it’s only sufficient to say that you can tell the difference between *the things in the test* with 95% confidence, not that you can tell the difference between ALL FLAC’s and MP3’s? Or am I wrong about that as well? I was still under the impression that you needed a large sample size to be able to make that kind of generalisation.

  33. Nick:

    For tagging audio, I use Tagscanner. I’ve used to do it manually in Windows XP’s explorer, until I found it a couple of years ago. My mp3 library has over 10,000 tracks in it. It’s really great, packs a lot of functions into it and really makes the sorting job easier.

    http://www.xdlab.ru/en/

  34. Ben:

    Paul,

    Love your music — Schism is a one of the best pieces of electronica in recent memory. However, the fact you’re making such a stink over the hassle of FLAC is decidedly unprofessional. Yes, this is your blog, yes, you’re an indie studio, but complaining about delivering content to your fanbase does not become you.

    Instead of griping about having to encode and tag an additional music format, I humbly suggest you take an afternoon and write yourself a nice batch file (or Bash script) to automate the FLAC encoding process. There’s bound to be a command-line tagging utility as well.

    You guys just shipped Red and it’s entirely possible you’re burnt out, or at the very least fatigued beyond reason. Having done game development for years, I know how draining releasing a game can be. But please don’t blame FLAC for having to deal with an inefficient asset pipeline. The pipeline can (and should) be optimized until FLAC is as painless to you as possible. That’s just smart development.

  35. urbanhusky:

    ABX testing one specific track with mp3 vs FLAC and getting a result means that the result only applies to the specific context of the test. That means: the track itself, the used sound soft- and hardware as well as the individual participating in the test.

    If I’m not able to discern mp3 and FLAC with a specific track, then it is not possible to generalise and say that I’m unable to discern between mp3 and FLAC all the time – the same goes for achieving a different result.

    Both sides of the argument (“You cannot hear a difference”/”I can hear a difference”) have to keep this in mind and not jump to conclusions.

  36. Paul:

    @Ben

    Thanks – I’m glad you like the music!

    I don’t fully understand your point, though.

    I’m going to continue supporting FLAC 100% – my feelings about doing so make literally no difference whatsoever to my fans. I’ve consistently said that I am never going to change my support of FLAC.

    I don’t have to like it and I do have the right to complain about it, just as you have the right to tell me that I’m whining and that I need to shut up.

    I thought about this issue and I decided it would be more interesting to discuss it in public because:

    1.) It might change my feelings and opinions on it

    2.) It might create a provocative debate that would be interesting to other people. Ranting on blogs is fun – I think we may have become a little bit too cautious around some of the things we say and I’d like to bring some of the interest back.

    I’m a musician, not a developer. I think the last time I wrote a batch file was probably around 1996. The response “STFU and just get a better tagging process” is absolutely something I welcome – I was expecting people to suggest things to make that easier, which they have done.

    “Don’t blame FLAC for having to deal with an inefficient asset pipeline” – WHY can’t I blame FLAC for that, if it’s objectively harder to deal with FLAC than it is with MP3? I’d like to blame FLAC for the fact that ALL common audio players handle it worse than MP3. Also, have you ever tried to encode FLAC on Windows 7 using the standard FLAC front-end? That seems like something it’s justifiable to complain about.

    I hope you continue to like my music even though I can be fairly irritating.

    @Urbanhusky

    Yeah, I think we’re on the same page with this actually. I basically just 100% agree with all of that.

  37. StephenM3:

    My thinking is this: it appears that some people, with effort, can tell the difference, but most people with “untrained ears” can’t tell the difference. But it’s still good to provide something to satisfy those who can tell the difference. It may not be something that everybody can tell, but not everyone has the same ears!

  38. Archagon:

    First of all, I’d just like to say that I really appreciate you putting out your music in FLAC!

    You’re absolutely right in regards to quality. Double-blind ABX tests on reputable sites like Hydrogenaudio have proven that the vast majority of people can’t hear a difference between lossless/uncompressed and even 128kbps LAME-encoded MP3, and V2 (~192kbps VBR) is pretty much the highest anyone should go. Except for specially-selected problem samples, I don’t think anyone has been able to consistently tell the difference between 320kbps and lossless/uncompressed.

    (And people get REALLY ANGRY when they hear this. Unfortunately, the placebo effect is quite powerful, and it’s hard to admit to yourself that the difference you’re hearing doesn’t actually exist. )

    FLAC does have a few benefits:

    1. It offers CD-quality. I know in your case there’s no physical version so it doesn’t matter as much, but for releases with both physical and digital versions, if there’s no lossless version available, I feel I’m getting ripped off for buying an MP3 album for $10 when I can get a CD with objectively better quality for less. Even if it makes no difference in terms of how it sounds, I don’t like feeling like I’m getting less by paying more.
    2. If you’re re-encoding files, FLAC is definitely better. I know this isn’t so much of a problem anymore, but what if you want to convert your MP3s to AACs? There will most certainly be quality loss, probably audible. If you start out with FLAC, it makes no difference. This also applies to people who want to fiddle with your music in audio editors. (Not sure if you’re OK with that, but there you go.)
    3. I mean, really, it just feels good to have a “definitive” version of the music in your hands, even if there’s a 24bit/96khz version out there somewhere. (Don’t tell anyone!)

    So yeah, basically what you said: it’s a great archival format. But I’m inclined to think that kids in the future won’t really care about this as much as we do, if at all.

    As for everything supporting FLAC: personally, I think people should start shifting over to Apple’s ALAC (also lossless), now that Apple has opened up the standard. Most iOS devices and tons of players support it, which is already more than the number that support FLAC.

    Tagging is pretty easy on Windows with foobar2000. I think you can just Ctrl-C all the fields from the MP3s and Ctrl-V them into the FLAC files. There’s probably something similar for OSX.

  39. Archagon:

    Just want to comment on this:

    “I still don’t understand why someone would focus on the FLAC vs. MP3 issue with an album that is demonstrably lo-fi in its nature.”

    To me, at least, it’s about preserving a level of abstraction between the music and its output format. Like, just because a piece of music sounds shitty doesn’t mean you should compress it in a shitty way, because the shittiness of the compression is very different from the shittiness of the music, even though you won’t be able to tell. But I’m a programmer, so these sorts of things irk the analytical part of my brain. 🙂

    And if urbanhusky can consistently tell the difference between MP3 and FLAC, that’s bloody impressive and I’m immensely skeptical. I’ve never heard of anyone being able to do this.

  40. urbanhusky:

    @Archagon I have left my ABX tests for anyone to see and like I said, I grew up training to listen to details.
    I’ve also exposed some of my friends to the same tests and they had little to no trouble discerning between mp3 and FLAC – where I found it to be much harder. If all you’re doing is listen to the same over-compressed music in mp3 with your iBrand earplugs, then I wouldn’t be surprised if you had a hard time discerning between mp3 and FLAC.

    Personally, I prefer FLAC because my “mp3” players support it (which I chose because of their sound quality, support for folders and USB mass storage use). I wouldn’t mind seeing ALAC out there, because I could simply convert it to FLAC without any loss in quality.
    I’m also a bit sceptical regarding the achievable sound quality with all these iProducts. I’ve heard about a constant decrease in quality in the iPods because they were cutting corners…

    Most people also don’t seem to give a shit about sound quality as long as it has enough “oomph”.

  41. Lyxar:

    Been a while since i’ve been in contact with mode7games. I suspect you and me still cannot stand each other 🙂

    Still, on this specific topic, i think we mostly are on the same line.

    Some background about me in this regard: I’ve been a very active member of the “objectivist”-community in the past, even on the central community on which FLAC, replaygain, ABX-testing and foobar2000 (which made many of those things popular) were developed (hydrogenaudio).

    About 4 years ago, i left the community. I never totally did fit in, but originally did – i’d say “believe” – in their conventions and mindset. Over time, i grew out of it and looked at objectivists and subjectivists in a similiar manner, as how i look at theists and atheists (just stfu both of you already, mkay)?

    So much for the background. Now as for FLAC and your post: I consider the main benefit of being able to download something in FLAC-format…. NOT to LISTEN to it in that format…. but instead, that i can convert it to whatever format suites me best, with zero quality loss (since the source is lossless).

    The same is not the case for 320kbit mp3. This is especially noticable, if the “album” consists of seamless track transitions. Even though a transcode from 320kbps mp3 may sound indistinguishable from converting from FLAC, the problem is that on the gapless track transitions, the samples do not align on the transcoded file, and thus a “POP!” is introduced. This is because even though the transcoded track itself may perhaps be indistinguishable from the original, our ears have no problems whatsoever, with noticing whan one sample is at let’s say -30000 and the next sample is at +30000. “Psychoacoustic” audio formats (and encoders) simply do not consider such cases – they treat a single track as isolated from the other tracks, and thus do not care if the first or last sample, may be off – because as long as one listens to just that one track, everything seems fine.

    Of course, you may now propose “Well, okay, then how about you just don’t transcode, and listen to the original 320kbps file?”.

    My answer to this quite possibly may not satisfy you: 320kbps CBR mp3 to me is “bloated”. It takes up way more space than is neccessary for “indistinguishable from the original”. Basically, 320kbps CBR to me is a bit like a misfit: Too good for listening, not good enough as a source to transcode from.

    Soo, what then…. release the tracks at lower average bitrate? Then some people will in turn complain, that the quality supposedly is inferior, and that they supposedly can hear differences.

    My point is: As long as you release as mp3, it is impossible to make everyone happy. Everyone will have something to complain about, because it’s either “too bad” or “too good” or whatever. What mp3 however offers, is something that can easily be listened to by people who “do not care” – who just want to listen to the music and don’t care about all the techbabble.

    And by the way: You are totally right that mixing is like 1000x more important, than marginal differences in encoding efficiency. Heck, remember that chat between me and you, about the determinance soundtrack, and me saying that its a bit boomy and “hot” – and you then explaining to me that you did that kickass track on totally simplistic equipment?

    So, to summarize: The reason for me favouring FLAC as downloads, has more to do with usability and choice, than encoding quality: I can from the lossless source just turn it into whatever my favorite format and settings are, without you having to provide that.

    And yeah: Mixing – goddammit….. mixing…… with a lot of tracks nowadays, i wonder if something like stereo even exists anymore, or if now every mainstream track is now mixed maximally hot in mono, so that we can then buy all that stuff over again, for 5.1 speakers.