Slashdot Mirror
New "mp3PRO" From Fraunhofer, But What About LAME?
xenoweeno writes: "This Week In Consumer Electronic reports that Thomson and the Fraunhofer Institute are working on a new mp3 format they've dubbed "mp3PRO." Thanks to competition from e.g. Windows Media Audio, they're looking to get 128kbps quality down to 64kbps. Great, but what does this mean for projects like L.A.M.E., which has just recently freed itself from Fraunhofer ["regular"] mp3 code/patents? Back into the fray?"
Source alternatives (png, ogg vorbis, etc) they are always in reaction to an existing technology brought forth by a company who could afford the R&D costs on something more original.
;)
I think you are making a wild assumption. What I think you dont realize is you are using your 'capitalist-centric' rationale to relate these things unnecessarily. Its a chicken-egg problem. I believe that because the modern social organization is essentially the corporation it plays a dominant role in everything. While I'll agree that new tech usually appears under a Corporate moniker - it does not happen as a natural result of something this corporation provides.
Research is done by people -- these people have motivations (fame, food, sex, whatever) -- if they were _allowed_ to produce work for the public domain: they would. The technology would still be there -- there would still be innovation. We have allowed corporations (or those who can use them as tools to satisfy their own needs) to gain _alot_ of 'power'. That power is all around you - your City Hall is a corporate entity and acts like one, your employer is a corporate entity and acts like one, your civil groups, your soccer league, your grocer -- EVERYTHING is a corporate entity and acts like one.
What I would advocate is a legislative restructuring and balancing of (Capitalist) Corporate vs Civic vs Private 'power'. We have (unfortunately) allowed (Capitalist) Corporate entities to 're-organize' culture to their advantage -- I use the analogy of a virus being capable of altering conditions within its host -- this effectively precludes any action outside of this dominant social structure.
So, back to your original point: Don't believe that innovation, freedom, basic needs and all else that you require for sustenance would vanish if Corporations did - that somehow they have 'given' us these things: It is very untrue. Technology would be born and evolve quite well outside of the present (Capitalist) Corporate world - maybe more so, because alot of the unfortunate after effects would be lost -- we could spend a greater amount of our efforts on worthwhile things (technology, art, and other self-satisfying pursuits) if we could abandon the 'necessary evils' of the corporate world: Advertising, Marketing, BizHeadTalk, Consumerism, Needless Pollution, Self Rewarding Greed, Corporate Enslavement, Exploitation of the Underprivileged and other various crap - people should be able to live their lives without being _forced_ to participate in (in what I believe) to be a selfish, ill-conceived system to fuck other people. I believe it is better to share with your neighbour than to steal from him - but what option do I have outside of being a modern-life-martyr and starving?
Ok - im done preaching.
Let's not be silly and just say that a compressed format uses a compression algorithm, okay? And there is a difference between lossy and lossless compression.
By the way, I'm an engineer so I've read some literature.
Gamingmuseum.com: Give your 3D accelerator a rest.
To use Microsoft's Windows Media Audio format as an example (because they make the same claim as MP3PRO): One of Microsoft's claims on WMA is that you get better quality at half the size. This is actually more true than most people (especially around here) are willing to believe. I thought it was marketing BS until last week, when I decided to run an informal experiment.
Using MusicMatch Jukebox I ripped one song ("Finding Me" by Vertical Horizon, because it starts off immediately with sufficiently dynamic sound) into raw WAV, 128 and 64 kbps MP3 and 128 and 64 kbps WMA. Then I listened to sections of the song, comparing the same section in the five formats. I'm a vocal musician, and have a sensitive ear to sound quality, particularly in music (I know it's not as rigorous as that waveform analysis of MP3 encoders that was published here recently, but in the end it's how the music sounds to the human ear that really counts).
I was impressed with the fidelity of the 128 kbps WMA. In fact (and I was quite surprised at this one) the WMA was truer to the original than the same bitrate MP3. The MP3 was actually slightly dampened in the high ranges compared to the original. What was really surprising were the 64 kbps files. The MP3 at that bitrate basically sucked, because the sample rate was only 22 kHz (resulting in weak low ranges and muted upper tones), while the 64 kbps file sounded quite good, because it was still at a 44 kHz sample rate. Of course, there was a slight degradation (to my hearing) in the sound quality between the 128 and 64 kbps files (mainly a slightly "metallic" sound in some spots), but I suspect that most people wouldn't really notice it.
So, MS's claim of 1/2 the size, better quality is not strictly true in combination; it's more like same size == better quality, half the size == considerably-more-acceptable-than-MP3 quality at that size.
If MP3PRO uses an even better compression/interpolation algorithm than MS Research came up with for WMA, then not only will 128 kbps MP3s sound better, but 64 kbps MP3s will probably start becoming the norm for use in pocket MP3 players.
--Ford Prefect
Now people (heavens no, not me) can download twice as much Metallica for the same price!
And the immediate comment by Slashdot is "But what about LAME?!". Aren't we being just a tad narrow-minded here? What the article subject says to me is "Fraunhofer is developing something new, which is bad because we've just managed to legally use the last thing they did."
Fraunhofer developed mp3. Had slashdot been around by then, it would probably have considered that to be really, really bad news because of the license. But, yet, I can listen to mp3:s today. The development of mp3 wasn't a bad thing in the long run, and there's nothing that indicates that the development mp3PRO would be bad in the long run either.
Couldn't we try to be just a bit positive about new inventions and developments instead?! Even if the inventions aren't made by three happy hackers in a University basement?
Why do I say this when it's not in the LAME docs? Because LAME is open source, I needed it, and I hacked it in.
It's that simple. (I've tried to interest the coders in this but they don't 'get it': I think mastering engineers will be a lot more interested.) Let me put it this way- as long as there are mastering engineers and studio types willing to struggle with the code enough to look after _their_ interests, LAME will beat whatever Fraunhofer comes up with, because Fraunhofer is all computer geek researchers- I see them going for a 'one size fits all' consumer encoder, and this is totally unacceptable and wrong for professional use.
And encoding music into mp3 or similar lossy formats _is_ a professional task: in the field of music it is so competitive that anything a band or artist can do to give themselves an edge _will_ be done. If LAME operated by a real mastering engineer using serious reference speakers and adjusted to let the soul of the song through beats the quality level of mp3pro operated by the artist... then the guy using LAME will score more downloads, get more attention and money, sell more CDs when his music _sounds_ _better_ at the same bitrate, and the guy using mp3pro and coloring his sound with a preset compressor that isn't tailored for the individual song will _lose_.
This is such an obvious development, and it is totally favoring the 'open source' side of things, because there are not enough 'mastering engineers', even wannabes, to tailor a commercial product to them- yet the work they do can outperform what you'll get from preset compressors. It's not the easiest skill to acquire but if you can do serious mastering on CDs you have the skillset to know what to do when you have your dirty little hands deep in the internals of LAME.
I'll demonstrate using my music as an example: the top album, "Marginal Theorems", and the second most recent, "Wounded Skies", are all made into mp3s using LAME with this ATH level control and extensive mastering work on the low and high frequency cutoffs. Psy model is turned off and replaced with ATH masking only, and ATH is custom set to the needs of the track to render the reverberant field with the proper weight, where necessary. The high cutoff is a great deal more gentle than you'll usually see, so that the tendency of the encoder to grab at all the HF data it can get is mollified without seriously altering the tonal balance of the HF sounds- basically the highs are eased back in volume until they take a place in the soundstage that isn't over-forward, but retain their character. The lows are set with a combination of low cutoff and slope that allows a sort of resonant area to be moved up and down according to the needs of the track- sometimes extending well below hearing, sometimes moved up to add muscle to the midbass.
Go ahead and hear for yourself. You should hear how good the _256K_ mp3s from the same settings sounded! (I'm probably going to find a site that lets me upload those for high quality buffs or those who want a CD made- probably Ampcast.com.) Any mastering engineer could do this, but at the moment it is LAME-only because I'm not aware of any other encoder that lets you meddle with the guts of the encoding parameters that way. I asked about this for almost a year now and nobody had an answer, so I finally did it myself.
Anyone wishing to roll this into the main LAME distribution will get full cooperation from me... so far I haven't seen an interest. Which, cynically, I don't mind so much: _I_ believe in the principle of free software, but if people don't _want_ the hack by which I'm able to encode ambient information properly, fine: I'll just use it myself. I happen to think it is one hell of a secret weapon. I posted on rec.audio.pro about my modified Lexicon reverb and used 'Marginal Theorems' (128K mp3) as a demo of what I do with my Lexi and people were _floored_. They didn't neccesarily figure out how much of that I owed to free software and LAME, though ;)
muahaha, free software as secret weapon!
You can't have it both ways.
I thought people believed that open source was better than commerical software.
So MP3 comes out, and open source can faithfully reproduce it by violating these (non-obvious) patents. It matches implementation but certainly doesn't improve on the efforts of Fraunhofer institute, their real work being in developing the acoustic model. If Fraunhofer can improve on that it just shows that potential for improvement was always there, but open source efforts weren't good enough to find it.
The only reason people would switch to this new encoding would be if it was substantially better. If that means that open source software falls behind - tough shit, this can be fixed eventually, it just means ignoring a different set of patents. This just shows where the real innovation comes from. I know that innovation is a dirty word now that MS have got their fangs in it, but there is such a thing as the genuine article.
http://rareformnewmedia.com/
adolf writes:
I don't -want- to see hear bad they can make things sound at 64kbps. And further, I don't care about bandwidth or disk usage, even though I'm behind a 28.8 modem.
I -do- want high-quality downloadable (freely or not) music. By high-quality, I mean indistinguishable from a CD to my own ears (LAME at ~220Kbps average VBR does this for me).
Storage is stupid-cheap these days. Bandwidth is slowly spreading out into much more diverse, and usually competitive, markets.
Sounds like what you are looking for is FLAC (http://flac.sourceforge.net). It's lossless compression. If you sample at CD resolutions, you get CD quality sound (if you sample at higher resolution from a better input source, you get better than CD quality sound). Only two downsides:
* Less compression than mp3's
* Not finished yet (but they do have working code)
Check it out.
----
----
Open mind, insert foot.
I was at a reverse engineering conference and happened to sit down next to a guy from the Fraunhofer Institute. They do a lot of research in many different fields and when I asked him about patents he shut up for a minute and then said "yes, we are commercial oriented, that's how we get to do so much interesting research." He didn't seem too happy about it.
How we know is more important than what we know.
LAME isn't "free from Fraunhofer mp3 code/patents". They may have finally outgrown their name and become a full-fledged mp3 encoder in their own right, but no matter, Fraunhofer's patent still stands. LAME infringes on that patent.
From the Vorbis FAQ:
(note that this question isn't on the faq from vorbis.com, it's from xiph.org.
No matter how hard LAME tries, it is another MP3 encoder, and as such, infringes on mp3 patents.
Higher quality closed formats is not the answer. Higher quality open formats are the only way.
Jeremy
Looking for a Python IRC bot?
Vorbis is GPL/LGPL too, which is a definite plus to many geeks :)
Bill Stewart
New Fast-Compression-only CPR http://preview.tinyurl.com/dy575ks
if they claim they can heve the same quality in half the data, does this also maen they have better quality in the same ammount of data?, e.g. is this going to improve the quality of a 128 kbps compression?
Sig (appended to the end of comments I post, 54 chars)
Many people seem to overlook VBR mode, i have yet to find a player that doesnt handle it, the rio, xmms, winamp all handle it fine (the bitrate meter goes silly, but hey) and you can get much better quality for the space, as it ramps the quality up and down as required, so your not wasting a few hundred kbps on the silence.
According to this http://www.neteconomie.fr/news/infoCOM.php3?id=838
(in french can't find anything in english, sorry), MP3Pro shouldn't be the only new codec blooming in spring 2001. Universal Musics wants to launch his new codec: BlueMatter (developped by Entrust (http://www.entrust.com/?).
According to this interview (once again, in french sorry) of the director of Universal Music France, BlueMatter should be used to make people pay for online music (I also read about Universal projects of online music and it seems to be streaming only).
So I guess that the new formats won't be as public as MP3 has been to prevent unauthorized players and encoders. One can always try to revers enginer the codec but it'll be hard both technically and legaly (especially in USA with the DCMA if they intermix an access control process with the codec). Beside, this was the strategy used by Apple with the Sorenson codec and unfortunatly there is still no free (as speech) Sorenson codec.
For my Ph.D. research I work a lot with audio codecs and the statement that they want a 64 Kbps bitrate to sound like 128 Kbps MP3 is doubtfull. They claim the MP3PRO format to be downwards compatible, the MP3 standard does not leave any room for a 50% reduction without a giant breakthrough.
A new technology is needed such as sinusodial coding.
MP3PRO Open technology? also doubtfull.
Johan.
Regardless of the philosophical implacations of the GPL, you *can* sell software that you give the source away too.
You're a programmer; you sell your service to write code, for whoever will buy your program, whether it be a business, end user, or whatever.
As a personal preference, you have every right not to want to give away your source code. But there is a distinction between selling software and writing source.
Software has been tested, debugged, packaged, polished and marketed.
Source is just that, source, like the distinction between architectural blueprints and a house.
If a GPL based company actually wanted to buy your source, is that any less or more a transaction than buying the software? The intent of some people, as regards Open Source, is that when you buy the software, you can get access to the source, otherwise you're relegated to a 'service', of providing support and fixes and improvements to a program, where the purchaser cannot fix, modify, or tweak the code to their taste.
That's just one view, at least.
Geek dating!
GPL Deconstructed
Not to sound pragmatic, but with any luck the new 'mp3' format will fizzle and die.
I don't -want- to see hear bad they can make things sound at 64kbps. And further, I don't care about bandwidth or disk usage, even though I'm behind a 28.8 modem.
I -do- want high-quality downloadable (freely or not) music. By high-quality, I mean indistinguishable from a CD to my own ears (LAME at ~220Kbps average VBR does this for me).
Storage is stupid-cheap these days. Bandwidth is slowly spreading out into much more diverse, and usually competitive, markets.
The focus should not be to make the files smaller (Realaudio G2, anyone?), but to make the quality better. The data distribution and storage capabilities of the Internet at large are progessing leaps and bounds ahead of the state of human hearing (which is actually moving BACKWARDS due to higher levels of everyday ambient noise) - once the epitome of perceptually perfect encoding is deemed possible for the masses, I'll settle for smaller files that reach the same end. Until that point is reached: Fuck off, Fraunhoffer.
And, dispite my freedom-esque views on life, I'd like to see high-quality encoding forced forced upon the populace, as the most infuriating members don't seem to mind even 96Kbps joint stereo mp3s either due to the fact that they are deaf, use equipment that is absolute shit, or just have never heard anything better.
It's for their own good, really - most illicit MP3s come from teens-to-20somthings who don't have to the cash to spend on quality (as in, "you can't buy this at Circuit City") audio equipment, but who (given the forward momentum of consumer electronics) will, at some point, be disappointed with the sound quality of the typical 128KBPS MP3 (of which they will have amassed several tens of gigabytes by such a point).
You idiots who bought a Diamond Rio (or similar) with only 64 megs, being pissed that you can only get an hour's worth of still half-assed-sounding music on the device, are no exception. You should have realized that flash memory is hideously expensive -before- you made such a purchase.
Feel free to moderate this down as flamebait. It's not like karma doesn't grow on trees.
Kid-proof tablet..