Slashdot Mirror

In double-bind studies, self-proclaimed "audiophiles" were unable to tell the differece between a 256kbps MP3 (CBR, encoded with FLAC) and the uncompressed CD.

Hydrogen Audio - many people there will say differently

Sure there is.. even you might notice if you knew what to look for and had a decent sound setup (good headphones, quiet computer, decent soundcard (i.e. not Creative)).

Chances are it sounds *different*, not necessarily much worse. You should try ABXing* with a lossless source and see when you stop being able to tell the difference. LAME is tuned to attempt to produce such a "transparent" file using --alt-preset standard, but obviously with such a limit format it can't work miracles (some music *really* needs more than 320kbps).

* ABX involves taking two tracks and distributing them randomly across (A,B) and (X,Y). You flick between them and try to match X,Y -> A,B. Thow in some stats, and you get a fairly scientific assessment as to whether you can tell any difference. On HydrogenAudio, you can get banned for making a statement about quality without doing this ;)

ah thanks for the heads-up. i hadn't heard of the evolution of lpac.

in any case, there's an informative thread on hydrogenaudio on the whole subject. suggests that it *isn't* mpeg-4-als, since it's as yet unfinished.

Of course, there are those with uber-expensive audio equipment that will tell you MP3 is inferior to Ogg Vorbis outright. I've used all sorts of LAME settings to get something comparable to Ogg Vorbis at 224 kbps or so, but why should I have to go through all those hurdles when I can simply encode my music with -q 7 to get something that sounds about exactly the same?

Besides, not all parameters work best with all sorts of music either - MP3 royally screws up a lot of recordings I have at any bitrate or option I've used (Profanatica demos, early Emperor recordings, Clandestine Blaze, etc.), but it's the same with Ogg Vorbis. So the only way for me to enjoy those without resorting to the originals is to encode it using FLAC or some other lossless compression.

The point is that encoding music "right" is an art that's difficult to get right since "one size does not fit all."

I just want to be able to play my songs whereever I want, whenever I want.

All of my iTunes (around 200+ now) are on a portable hard drive. I should be able to take that drive anywhere, plug into my laptop.. my friend's PC.. my work PC.. etc. and play the songs without having to "authorize" a machine.

Nor do I want to have to deal with installing an application (read: iTunes) that has a horrible/resource-inefficient UI (brings my 1GHz laptop to a crawl) when there are much better designed and implemented players out there that can run standalone, and do not need Quicktime or iTunes installed.

I'm not a pirate, and I wish companies would learn not treat their customer base like they are.

There is a winamp visuaization bridge that works with milkdrop called vis_bacon.

Binary

official forum entry

Ogg Vorbis has been shown to be non-transparent in a number of situations all the way up to -q10 as a result of issues with preecho and the like (with the standard reference encoder at least). Even MusePack, widely regarded as the most well tuned lossy audio format around, isn't perfect in all situations. As for MP3, well.. if Vorbis and MusePack have to hit 400+kbps to remain transparent on some tracks, MP3 sure as hell can't hope to do so with it's limit of 320kbps; the very format itself is inadeqate.

Even ignoring that, there are plenty of other reasons to go lossless; transcoding and such are *much* better with a lossless source; doing so from a Vorbis or MP3 file is almost certainly going to produce artifacts you can ABX. With a lossless file, I can convert to any format I like as many times as I like, since I effectively have the original.

As for the extra bandwidth consumed; well, not everyone values quality as much as I do - it's not such a big deal if one in 1000 of your users uses a few times more bandwidth, not to mention that those users are likely to be willing to pay a bit extra for it. Even if they're not, bandwidth isn't exactly *that* costly; the difference between downloading a Vorbis file and a FLAC file is going to be a couple of pence at most. If your margins are that slim you might as well give up now.

http://www.minidisc.org/atrac3.zip

You can use it in an application like CoolEdit or VirtualDub. RealProducer has an encoder as well. I recommend posting in the forums at http://www.hydrogenaudio.org if you have any further questions.

Well, seeing how bad ET's iTunes Bad, WMA Good article was, I figure Doom9's codec comparison is better than this.

And yes, Doom9's comparison includes XViD.

For good reasons, the posters on Hydrogenaudio don't take kindly to people making unfounded assertions about which codecs are better, so if you're going to argue with them, think twice and ABX first. You will be, after all, arguing with many audio developers, e.g. people who make contributions to LAME, people who've tuned the Vorbis encoder, and a surprising number of people who work for Ahead (makers of Nero, of course).

Winamp 5.02's encoder (which got a lot of help from hydrogenaudio's own Menno, a FAAD AAC-decoder developer/Ahead MPEG4 developer) wasn't included in the listening test because of a bug they found before testing.

Too bad, too. I would've loved to have seen how it compared.

hydrogenaudio.org

Don't be so lazy :D

--Quality encoding, even if it is VBR.

wtf?
VBR is the only way to go for a high quality/size ratio, and the only settings that would be better than --alt-preset standard is preset extreme (which doesn't really gain much in quality, only size) or just flat out 320kbps (preset insane). both these settings would add a lot to the file size.
and, yup, lame 3.90.3 is best lame version for use with the presets. so they absolutely know what they're doing.
this is as good as it gets for mp3, if they want to do any better they'd have to encode with flac...

btw, take a look at the forums over at hydrogenaudio.org

Wow, I am impressed as well. According to the 'experts' (I don't use the term lightly since the people there surely DO know what they're talking about) over at Hydrogen Audio also reccommend those LAME version/settings over the newer 3.93+ versions for higher quality encodes.

MP3 was won.. long before WMA appeared. It offers transparency on all but a few special samples at around 200kbps, and with storage getting cheaper, slightly more efficient codecs (Ogg Vorbis, for example) don't offer enough of an advantage for most people to move. I won't touch WMA with a long barge pole.. just because you made the (mistake IMHO) of going over completely to it, doesn't mean anyone else has to. Go read some very informative discussion at Hydrogenaudio.org for specific technical reasons not to use WMA.. other than being from Microsoft etc. Of course, there is a danger that many people will use WMA just because MS make it easy for them to get into it... but why that's a reason to advocate WMA, i can't imagine. It's unlikely MP3 support will be dropped in hardware any time soon I think... i'd be more worried about your sound quality and portability of those WMA files.

128kbps Listening Tests Results on HydrogenAudio:
Link

Nullsoft's privacy policy is listed here.

Why would you want to use your audio player to encode, and encode to mp3s at that? Get the right tool for the job, CDex to make ogg, flac, or even mp3 if you so desire. And you can go to RareWares to obtain updated dll files for encoding via CDex. I don't see any problem with a Pro version so long as necessary functionality from a media player isn't taken out.

foobar has native id3v2 support since 0.74

While there probably isn't a be-all-end-all codec pack for your specific OS

Actually, there is -- for Windows at least. But I've tried it before, and installing every possibly codec, most of which you never use, only makes things worse, not better. I find it better to ignore the all-in-one packages and just get the codecs from the proper sources. So grab xvid and divx and ac3filter--that'll cover most of it--and then grab other ones as you need them.

what web resources are useful for people who need to find a codec (even an obscure one) before they can play their media files?

Which is where AVIcodec comes in. What it does is show you what audio and video codecs are used by your media files. So when you find a file that doesn't work properly, it'll let you know what to look for. Google is normally enough after that, but RareWares is a good place to find some of the more obscure codecs.

...and even non-nogap mp3s (the Karma will now drop silent frames starting and trailing mp3s)

1. Use an encoder that stores two important variables, encoder delay and encoder padding, such as LAME 3.90.3 (recommended version.)
2. Play back MP3s that knows how to use these values to achieve perfectly gapless playback - I only know of foobar2000 that does this at present.

I am the person who did the original testing for MacRumors. Here are the final steps:

The raw aac file that QTFairUse produces can be played in a windows app called foobar.

To play back in itunes is a little harder. One must run an application called faad.exe to fix the "atoms?" of the aac file. After that is done one must add the MPEG-4 wrappers using the program mp4creator found in MPEG4IPutils. Make sure to use the -optimize tag, or else the file will triple in size. After this is all done you end up with a m4a file with the decrypted aac content in a MPEG-4 wrapper playable in itunes.

thread here.

People have recently been asking me what software I use, so I've decided to make a list of the software that I think is "essential" and that I have to install on any system that I use; feel free to discuss. Please keep in mind I'm only focusing on Win32 software here, since this is probably the OS of choice for most forum readers.

I've decided to go through and update this list to reflect my current configuration, and I'm going to include non-Win32 software, but only that which I use on a daily basis. This update should be complete by August 7th.

I intend to add quite a few programs to the list, and some are getting removed.

A sample of programs I'm going to add or expand on list: Subtitle Workshop, mp4ui, LA (lossless audio), ffdshow, AC3Filter, Zoom Player, The Font Thing, XnView, GhostScript/GhostView, DiscJuggler, COM Explorer, ActiveX Explorer, ZTree, BHODemon, VSigGen, PerfectDisc, PowerBASIC 7, 7-Zip, QuickPar, SFV Checker, vile, Borland CodeWright, Visual SlickEdit, Eset NOD32 Antivirus, WinSCP, AvantBrowser, lftp, NFTP, NcFTP, Sam Spade, Xnews, Mulberry, PuTTY, OpenChat/32, C-Kermit, Kermit-95, Open DCL, foobar2000, bfilter, Zsh, wget, ScrollZ, OpenSSH, Seminole httpd, and more.

I'm going to add a small section for NeXTSTEP and OpenStep software as well, since I'm a daily user of NeXT workstation equipment.

How about FOOBAR2000?
fb2000
and it is very extensible, very customisable, and has very low footprint. The author has written plugins for something, I can't remember.

.. and the results are here. That news was submitted before to /. but didn't make it through.

RTFA. VBR was used for Ogg Vorbis in both test.

There are audiophiles, and then there are people who really can hear the difference. I'm talking about people who do extensive ABX blind testing, never accept information or opinions not backed up with scientifically-acquired evidence, and who, in general, really know their stuff when it comes to audio compression (being ones who are actually coding these encoders).

I mean, seriously, I'd call these guys audiophiles, but the word has already been defiled.

Sorry to burst your bubble, but you might wanna visit hydrogen audio. They kinda know more about it then you or I. They do alot of extensive testing and many developpers of LAME/Ogg Vorbis/mpc/aac/speex/flac and others are there. They also tune codecs to optimize them. I wanna keep the post short so I''ll stick with this conclusions that the folks over there have: (My apologies if I make a mistake) MP3 is the old genertion. It's quite good if you use LAME, --alt-preset standard (average bitrate with this setting is 180-220 K) is transparent on most samples. But the next generation codecs are AAC and MPC, of which AAC is endorsed by many big companies while mpc is more of a personal hackkerthing made by enthousiasts. They perform much better then MP3 and lossy audio compression to the limits. Downside of AAC is that it's still young, which is actually an upside as this technology may surpass MPC, the current overall king, when it is matured. There is an encoder race going on. Ogg Vorbis is also new and next gen but unlike mpc/aac it isn't optimized for high bitrate's and it possibly will never surpass AAC/MPC at the high end.

I may be late to add this, but HydrogenAudio has just begun the second testing phase. See the announcement page. "Building on the AAC at 128kbps test results, this test is meant to compare the AAC winner (QuickTime) against other popular compression formats: Musepack, Vorbis, WMA Pro and MP3." Submit your results by August 3.

Actually, one of the purposes of this test was to weed out the best AAC encoder for participating in a future test that will compare it to MPC, Vorbis, MP3, and WMA. The discussion of this test can be found over at the Hydrogenaudio forums.

I used to like Winamp before AOL took it over.

I've had more problems with Winamp 3.0 that it has totally destroyed my whole view on the winamp product.

I found a program called "foobar2000" and I like it as much as I did the old winamp. It lacks all those teenie bopper popular skins, but underneath a very ugly, so simple yet confusing interface, it just sounds better then Winamp.

The key here is to change Buffer-ahead on track change to something higher than 0 ms. This means that as the end of a song approaches, Winamp starts to decode the next song so it can seamlessly play it as soon as the first song is finished. If it's set to 0ms, then it can't do this and there *will* be a gap between tracks - I, and many others, can certainly hear them.

Since this is getting rather off-topic, I'll end the thread here but I suggest you check out Hydrogen Audio, a very good site for almost anything high-quality audio. Just remember to use the search and find out what you can before you ask any questions to avoid getting flamed.

Cheers.

See for yourself.

Most of the really good Winamp coders left already. The guy that wrote almost all the core plugins, Peter Pawlowksi, quit because he didn't like Winamp 3 design and thought it was a dead end. Because AOL still owns his code, some of the plugins are 'dead' now, and the code can't be used any more. Bummer.

He wrote his own player instead, which is, eh, quite different from Winamp, Foobar2000.

Anyway, Frankel has little to complain about. Nullsoft was bought out for almost 86M$. For that much money, he'll never have to code, err, express himself ever again.

You should not use --r3mix. It is old and deprecated - its removal from LAME has been considered. You should use LAME 3.90.2 with --alt-preset standard (aka "APS", ~ 192kbps VBR) or possibly --alt-preset extreme ("APX", ~ 256kbps VBR) for trickier encodes (classical, jazz, rock, experimental). Those without space concerns still wishing to use mp3 can try --alt-preset insane ("API", 320kbps CBR).

The --alt-presets are optimisations for quality and have been very thoroughly tested by hydrogenaudio. They represent the current state-of-the-art in mp3 compression.

For a scale, quality (normally transparent up to lossless) and size (50-80MB up to 300-700MB) go roughly (Qx represents Vorbis 1.0 quality number): APS < Q6 < APX < Q7 < Q8 < API < Q9 < Q10 < FLAC

A music sharing network for people who care about quality exists. Because the bad guys read /. too, I'll leave it to you to find üs, but the rules are:

Rip with Exact Audio Copy 0.9b4 (secure mode, accurate stream, NO C2, no normalisation, no read or sync errors, only complete discs with no missing audio tracks, save a log file) and encode to MP3s (LAME 3.90.2 or 3.92), Oggs (Vorbis 1.0) or FLACs. Tag correctly - for mp3 ONLY use id3 v1.1 and id3 v2.3.0 - with year and ideally genre from allmusic, name scheme "%A - %C\%A - %C - %N - %T" normal, various artists discs - name tracks "Artist / Title" and use name scheme "%C\%C - %N - %A - %T", add " (OST)" to album name for soundtracks. Move log into directory, rename to directory name + .log, add an .md5 md5sum for the log and audio files to complete the rip.

Gar. I guess that's what I get for trying to make a post while I'm having trouble even focusing my eyes.

Anyway.. continuation of the third paragraph:

At high bitrates, Vorbis starts losing its advantage, not because it sounds particularly bad at high bitrates, just that it doesn't really sound any better compared to lower bitrates. MP3 on the other hand, goes from unlisten-to-able at 96kbps and lower (for your average user probably.. 128kbps is BARELY listen-to-able IMO) to sounding fairly good at 160-192kbps.. to sounding fairly transparent at the --alt-preset standard level (which, IIRC, ranges from 180-224kbps).

Among the only real "audiophile" community I know of, Hydrogenaudio, the "in" thing at the moment is actually neither Vorbis nor MP3, but Musepack. It does have the same possible patent problems MP3 has, but I really do think that MPC is probably the de-facto high quality music compression algorithm out at the moment.

For archive purposes (backing up music CDs and such), if you have the CDs for it, I'd skip the lossy codecs all together and go for APE or FLAC.

For play on the harddrive, unless you're an audiophile with great speakers/headphones, I'd say a mid-quality level ogg/vorbis encode would probably have the best bang for your byte, even more so on a MP3 player with limited space.

Wrong wrong wrong wrong. Wrong.

For a long time now, --alt-preset standard has been *the* premiere lame switch. --r3mix is just an alias for some other switches. --alt-preset is its own switch (ie, it tunes some internal parameters that others don't). --r3mix is now obsolete. The lame devs hang out at Hydrogen Audio, so you can go there and see for yourself.

The r3mix myth needs to die.

It looks like MP3s. Which is good. Is it all really just encoded at 128 though? Any option to get something better?

This discussion seems to indicate that they're moving their service to providing fairly decent LAME encodings.

I understand XviD's implementation of MPEG-4 is based on H.263.

XviD follows the MPEG-4 ASP (advanced simple profile) spec. Virtually all of the current major video codecs out there use some minor variant of this.

H.264 usually refers to the MPEG-4 AVC (advanced video coding) profile. This promises a 2-4x size improvement at similar quality to the ASP. However, it has one major problem...


So is anybody (including XviD) considering implementing it? I understand it isn't patent-encumbered. I could be wrong...

Yes, an AVC implementation exists, but it provides its own demonstration of why no one uses it yet despite the improved size and/or quality... Namely, 30-45 seconds per frame at encode time. For a full-length movie, that comes out to two or three days for a single-pass encode.

Additionally, even if you feel inclined to wait that long for the sake of quality (personally, I would), the link I gave above points to more of a proof of concept than a "real" viable codec. It needs quite a lot of tweaking just to make it compare to existing ASP codecs such as XviD.

Here is a rather amusing message I saw over at hydrogenaudio.org about AAC support on the Nomad series.

Try some of these and see what you think.

To do a true test, you need to encode the files, decode them to PCM wav format, then burn to an audio CD.

Then, you have to do a blind test with all of them. You also need to use a variety of source material, because different genres of music compress better under some encoders.

Learn why you shouldn't use spectral analysis to determine lossy codecs' quality.

The most respected technique is double-blind testing using an ABX tool such as PC ABX, WinABX or ABC/HR.

More info on conducting blind tests can be found at the PC ABX site.

Mp3 doesn't have to mean lower audio quality. A lot of tests have been done by audiophiles and mp3's encoded correctly are indistinguishable from the wave files even for most audiophiles. In a lot of cases mp3's are better than ogg's as the LAME mp3 encoder has been tuned at high bitrates to ensure good audio quality while ogg format is only now being tuned at high quality settings. See hydrogenaudio for info on various codecs, chrismyden for info on how to create high quality mp3's and Ubershare for info on how to share your high quality mp3's, ogg's, MPC's with other people who only share high quality files. And until there are some descent harddisk players with ogg support most of us will keep trading mp3's because they are more useful. In the only real advantage that makes me want to use ogg's is the fact that they support gappless playback, which is still lacking in all the harddisk mp3 players.

Arguably the best resource for audio compression information can be found at Hydrogen Audio. Visit the various forums, check out the excellent Foobar2000 win32 multiformat audio player, and learn.

I have also written a guide on ripping high-quality MP3s using CDex, aimed towards beginners. If you know people who use Musicmatch, help them switch to a decent, easy-to-use CD ripper.

Cheers,
CD

Arguably the best resource for audio compression information can be found at Hydrogen Audio. Visit the various forums, check out the excellent Foobar2000 win32 multiformat audio player, and learn.

I have also written a guide on ripping high-quality MP3s using CDex, aimed towards beginners. If you know people who use Musicmatch, help them switch to a decent, easy-to-use CD ripper.

Cheers,
CD

But not Emmett Plant I guess...

Sadly, r3mix is outdated and kind of obsolete by now. The alt-preset modes of LAME (as noted elsewhere) are recommended now. One of the web's centers of encoding discussion is Hydrogen Audio. They have very good FAQs available on the topic of LAME tweaking (although the conclusion is usually "use alt-preset-x). They're also an awesome resource on anything else related to PC audio, and they host the best Windows audio player available right now, Foobar 2000.

Sadly, r3mix is outdated and kind of obsolete by now. The alt-preset modes of LAME (as noted elsewhere) are recommended now. One of the web's centers of encoding discussion is Hydrogen Audio. They have very good FAQs available on the topic of LAME tweaking (although the conclusion is usually "use alt-preset-x). They're also an awesome resource on anything else related to PC audio, and they host the best Windows audio player available right now, Foobar 2000.

I disagree about MP3 being a "not hugely high quality codec".

MP3s encoded at 128kbps CBR (constant bit rate) using an encoder such as Xing WILL result in poor-quality mp3s, easily discernible by the averagle listen using poor quality equipment. However, an mp3 encoded using a recent version of LAME (i recommend 3.90.2) and "--alt-preset standard" will find that the resulting files are virtually indistinguishable from the source CDs (even to audiophiles), at an average bitrate of around 192kbps. This is superiour compression to ATRAC, and the LAME psychoacoustic model is significantly better tuned IMHO.

For more information on ALL lossy and lossless codecs by people who really know their stuff, check out the message boards at Hydrogen Audio.

Not true! While this is true for the Xing encoder which a lot of bad CD-ripper applications use, LAME or the Fraunhofer mp3 encoders do not mangle the sound as much. Try and get a good compile of LAME from HA and do a "lame --alt-preset standard file.wav" and open the resulting wave file in Cool Edit and have it show you the frequency stats of the created mp3. You shouldn't use this for a comparison of sound quality though - your ears are still the only useful tool for that, but what it will show you that you are wrong.

You might have heard of the Nyquist theorem before, which states that when storing audio data using PCM (Pulse Code Modulation) the highest frequency you can have it half the samplerate. Thus there are no frequencies higher than 22.05kHz on a CD (because their samplerate is 44.1kHz).