Programmers Will Debut Free MP3 Alternative

An Anonymous Coward writes: "According to an article on CNET.com, a group of programmers at iCast have created a free alternative to MP3, named Vorbis. According to the article, they're planning on showing a beta of Vorbis "at next week's MP3.com summit in San Diego," that it will be released without IP restrictions, and that it will provide equal or better quality than MP3. Gotta love free software!"

Comments

The skinny yo.

[jon_c]

As someone else mentioned this has OggVorbis has already been "new"
http://slashdot.org/articles/00/ 04/11/118219.shtml

Since then I have taken the time to actually check it out; compile the winamp plug-in, compile the encoder, browse through the mailing list archives

First off the code strikes me as very clean and well written, It looks like the guy knows what he's doing. Second he sounds like he knows what he's doing, he talks to people about the idiosyncrasies of audio compression, DSP etc... so I defiantly give the author props.

As for "how good is it". Well the skinny is that it's a little bigger then mp3, and a little lower quality, also encoding a 5 minute song a PII 500 took around 1/2 an hour. However REMEMBER IT'S 1.0. windows 1.0, Gnome 1.0, all sucked. This doesn't suck. And it's not even 1.0.

The author expects the low bitrate compression to surpass mp3, it's just a matter of time to get things finished.

Right now it looks like 90% of all the min-projects are done, they just finalized the bit stream format, xmms/winamp plugs only miss streaming support. And the command line project is nearing completion. Next I'm sure they will attempt to optimize it, and tweak the audio quality.

Somewhere in the mailing list I noticed the author was talking about how he kept the specifics of the quantization process open. Meaning it could be changed very easily, which in turn means that the compression could be very precisely tuned, that should be much more useful then simply picking bitrate/hz/stereo.

OggVorbis has the smell and feel of a next generation audio codec, It's open source, free and not owned by patents. I can't wait..

-Jon

Re:The skinny yo.

[Chris Johnson]

"Somewhere in the mailing list I noticed the author was talking about how he kept the specifics of the quantization process open. Meaning it could be changed very easily, which in turn means that the compression could be very precisely tuned, that should be much more useful then simply picking bitrate/hz/stereo.

Now this is what interests me. Usually when I am participating in discussions on this kind of stuff, it is as a musician, but underlying that is more than ten years of extremely hardcore audio geekness- I would feel pretty confident in saying that w.r.t sound engineering hardware I am as good as Monty is with this compression programming. (I realise that sounds like a big claim- it is. I make no such claim about programming or software :) )

From that viewpoint, I have to say that the idea of being able to 'very precisely tune' the compression is wildly exciting, just wildly exciting. Does everybody fully understand where that leads? It gets away from the computer geek core skills, but you know there are geeks whose core skills it leads _towards_. What you are talking about is effectively a form of MASTERING. Let me take a minute and give a little background on what this could be like...

First, mastering. Once music gets mixed down to a final two tracks (or more, if you're doing funky surround stuff, but mostly it's two tracks), what you have isn't the final media. You have a 30 ips openreel two-track tape, maybe, or you might have a 48k/20bit digital master if you're smart, or possibly just a 16 bit 44.1K DAT or something if you're not. If you have a CD you've basically done the mastering yourself- and that's not necessarily good, because you might suck at it and the idea is to do it really really well, possibly at huge expense and possibly not.

Mastering engineers always desperately want some form of much-higher-quality media to work from- if you are making a CD they would prefer 20 or 24 bit output. If you're making a record, you could use that or the traditional two-track open reel tape which generally exceeds vinyl playback quality levels by a wide margin. In some cases a mastering engineer making a CD might want analog open-reel tape instead of _any_ form of digital: the idea is that he knows waaaay better than most musicians how to do the transfer, and may have lots of sophisticated equipment to extract every bit of detail and color off that tape and make it jump out of the CD sounding totally lifelike. Mastering engineers are the ones who will get mad at you if you go around normalising all the levels on your digital master for 'em- the mastering engineer tears his hair in angst at such a situation because he or she knows exactly what damage the algorithm is doing to the sound, and that it can't be undone once done. In analog the mastering engineer handled compressing the sound and imperceptibly summing the bass frequencies of each channel to utilize all of the LP's ability to put across music. In all-digital, for instance the example of mastering a 48K 24 bit mix to CD, the mastering engineer will be keeping untouched versions of the files and experimenting with various EQ algorithms, different multiband compressions, all to try to FORCE the resulting lower quality version to sound BETTER than the supplied master tape. That's really the name of the game- making it sound as good as, isn't enough. And that's where you get into really expensive equipment and total voodoo engineering and guys paid on the scale of computer professionals whose names are recognized by lots of studio engineers and almost nowhere else. Ever heard of Bob Ludwig, Wilma Cozart, Bob Clearmountain?

Now. To bring this neatly into context with Ogg Vorbis, imagine getting access to a Vorbis 'compression settings box' that took up the whole screen crammed with little settings and adjustments. Stuff like 'masking threshold curve for left channel's frequency', or whatever goodies are there- the stuff that apparently none of the MP3 encoder writers think is the user's business. Almost everyone in the world would take a look at that and run screaming. Almost all computer geeks would look really confused and wonder, what is the point, isn't there just one right setting? Almost all serious mastering engineers would take one look and cry YEAAAAHHHH GIMME MORE O' THAAAAT! ...and they'd be off and running, and you wouldn't believe what they'd be capable of. _Every_ _single_ _song_ would be tailored to the optimum compression details for THAT SONG, not some mythical standard...

From what I've seen, encoder programmers seem to think the idea is to match test tones as accurately as possible- and that is not going to be good enough, going into the future with digital compressed media. It took reading that comment about 'very precisely tuned' to realize just what's happening with all this. Look, if you told a big recording studio or an 'audiophile' act like Pink Floyd that they had to have their music mastered in just one particular way, but that's OK it's 'optimal', they would KILL AND EAT you *g* the fact is, recording projects vary so widely in both recording quality and SOUND that it's totally, totally impossible to 'master' to mp3 or anything else using a preset algorithm no matter how clever. Compare the sound of Rush (heavy on pushing the edges of the frequency range, dynamic, over-clear) with the sound of Tangerine Dream (really deep ambiences, liquid sound _necessary) with the sound of AC/DC (punchy as hell but it's gotta hit you in the stomach, no shrieky grating edges allowed and no show-offy super low bass) and you can see the problem. All compression must trade off _some_ qualities. But how can one arrangement cover all those bases? That's like mastering all their tapes through the same EQ and compression settings- disastrous.

I always think of a particular guitar note when I think of mastering- it's on Alanis Morrisette's "Jagged Little Pill" CD, I think on the song 'You Learn'. There's a brief instantaneous moment where an acoustic-sounding guitar, quite in the background, 'snaps' a tiny highpitched note on the high E and instantly mutes it again. Sounds like a set of light electric strings on an acoustic. This one short note cuts through the whole mix with stunning airiness, like a tiny instantaneous holograph of the note popping out. It's magical, impossibly delicate and clean- and that's what mastering can do for a song. That Alanis album is brilliantly mastered by Chris Bellman at Grundman Mastering- without him, that moment would NOT BE there. It might not have even felt that way on the original tape! The guy reached out and found the essence of the sound and caused the CD to cast exactly the right sonic spell to reveal it.

MP3ing that track almost certainly obliterates that moment and makes the little 'plink!' into a tinny clonk, ruining it. But if it was possible for the mastering engineer to also master the MP3- I guarantee that things would be very different. If I was doing it I would key off that sound and weight the 'masking' or whatever so that you had the deep pulse and a big emphasis on the airiness- the highs would be stealing lots of data for themselves. The thing is, simply going 'Oh OK' and coding in a radio button for 'more air' is unacceptable- if I had to master that track to MP3, I'd need to keep going back and adjusting the curves of the masking and all that until that instant and other important moments jumped out of the resulting mp3 JUST EXACTLY like it did on the original recording- or *gasp* BETTER! It's not at all a question of finding 'the masking thresholds of the human ear' and coding in those. Mastering engineers need total control over all those details at an amazingly low level, even though they are NOT programmers and wouldn't know memory management or b-trees if one bit them on the ass. Mastering engineers know better than the programmers how to set the 'equaliser dials' for the compression- because that is their JOB, to know that.

So _on_ behalf of all the other mastering engineers who don't read slashdot and aren't rubbing elbows with programmers as great and brilliant as Monty, let me just say- "Give us the controls!" We can't do your job, don't try to do ours! Every project has vastly different needs, and sometimes the needs are quite unusual! I could easily see, for instance, a song in which you masked 3-5K (normally the presence area!) very heavily and threw lots of data at the extreme highs which might normally be masked! I can even name a song that calls for that off the top of my head- "Hot Sun", from Adrian Belew's album "The Lone Rhino". Lots of Bowie's "Low" album would demand a _very_ different compression than most material because it is meant to sound claustrophobic, sometimes quite artificial and grating, and sometimes totally Enoed out lack of highs. I could keep on giving examples for ages- the point is, pleeeeease give us arbitrary control over these things and stop trying to perfect them for consumers all the time. There needs to be a way for mastering engineers to practice their craft on compressed media. If you make that possible, then we will start getting music from talented mastering engineers that really pushes the technology to the limit- and the difference will not be subtle, it will be a massive blast of very individualised sounds, letting different bands _really_ get their own distinct sound, and everybody will be able to appreciate that while still being able to use the 'consumer' version of the encoder (with all the scary parameters covered up and set to nice 'optimal' values) to rip metallica CDs *g*

If you can't force Ogg Vorbis to be the mainstream format with the help of mastering engineers, maybe could you make an MP3 encoder with a similar level of adjustment over its parameters? This needs to happen.

If I understand things correctly, MP3 encoding can be free (or partly?) but it's been hard for the free encoder coders to catch up with the totally patented and proprietary Freihofer stuff. And so you have Xing which gets tinny and artifact-y, and Blade which is said to be dull and lacking in extended highs, and LAME which some people claim is 'perfect' because it comes between the other two in personality.

When will people understand that these are all settings on the EQ dial? Mastering engineers do not need a single black box with the 'ideal' settings locked in painstakingly. Mastering engineers need KNOBS.

Give us KNOBS! :)

Haiku

[575]

Best compression rate:
Poet reduces music
To simple haiku

File extension

[rana]

Why use .ogg as the file extension. Why not an extension based on "Vorbis": .VBS

Oh wait, that one's already taken, it means "Virus Building Script".

Re:Still Interesting even if it's not news

[Bruce Perens]

User acceptance of audio formats is driven by one thing: publisher use of audio formats. Given that a plug-in is available on the net and easy to install, the user will download it the first time he has to play a file in the new format.

In this case, a lot of publishers don't want to pay the Frauenhoffer Institute royalties, and this new format is a way to get out of that. Nor do software developers want to pay license fees.

If we want to push it, the best way is to start writing applications for it, and to start producing audio programs in it ourselves. The Free Software community is an effective engine for driving early acceptance.

Thanks

Bruce

see this:

[LocalYokel]

http://slashdot.org/articles/00/ 04/11/118219.shtml

--

Oh, you need to read the hidden seekrit FAQ! ;-)

[xiphmont]

Yeah, that one isn't on the FAQ.

People keep asking, so I'll just post it. From the name page:

The Ogg project began with a few-weekend-attempt at a simple audio compression package as part of a larger project in 1993. At the time, the software was called 'Squish'. The project and the general problem of music compression became a personal fascination, and Squish took on a life of its own far beyond the proportions of the original digital music studio project of which it was to be part.

A few months after the first Squish webpage, I received a polite but firm letter informing me that Squish is a registered trademark (for a mail transport system). Mike Whitson, a contributor to the cause in the early days, suggested the name 'OggSquish' as a replacement.

An 'Ogg' is a tactical maneuver from the network game 'Netrek' that has entered common usage in a wider sense. From the definition:

3. To do anything forcefully, possibly without consideration of the drain on future resources. "I guess I'd better go ogg the problem set that's due tomorrow." "Whoops! I looked down at the map for a sec and almost ogged that oncoming car."

(see the rest of the definition for the original Netrek usage.)

At the time Ogg was starting out, most personal computers were i386s and the i486 was new. I remember thinking about the algorithms I was considering, "Woah, that's heavyweight. People are going to need a 486 to run that..." While the software ogged the music, there wasn't much processor left for anything else.

These days, Ogg is a larger multimedia project that does not only concern compression; Squish became the name of one of the Ogg codecs. For that reason, we usually just refer to it as Ogg when there's no Netrek context nearby. The Ogg project has nothing to do with the common surname 'Ogg'. Nor is it named after 'Nanny Ogg' from the Terry Pratchett book _Wyrd Sisters_.

The 'Thor-and-the-Snake' logo is drawn somewhat from Norse mythology; the real symbolism is the sine-curve shape of the snake. Thor is hefting Mjollnir about to compress the periodic signal Jörmungandr... See, it all makes sense.

Vorbis, on the other hand is named after the Terry Pratchett character from the book _Small Gods_. The name holds some significance, but it's an indirect, uninteresting story.

Monty

Re:Filesize is King

[drix]

First, the software is pre 1.0.. to quote another poster on this same thread, "REMEMBER IT'S 1.0. windows 1.0, Gnome 1.0, all sucked." I couldn't have said it better myself. It's not fair to judge this guy's work based on something he hasn't technically even finished yet.

Second, and this is why you should be damn impressed, one man created this entire algorithm himself. That, my friends, is a really, really hard thing to do. Fraunhoffer had a think-tank stocked with well trained engineers for them to come out with MP3 the algorithm, to say nothing of making the crucial leap that a lot of sound sort of "cancels out" in our head and may therefore be ommitted - that doesn't even strike me as in the same field as computer science. For this guy to even come close to rivalling those achievements, alone... well, ever hear of the Small Pond syndrome? It just reminds you that there are people out there that are a lot, lot smarter than you :) Read his technical discussions available on the sight regarding wavelets, DSP, etc. and you should be suitably impressed... I was.

I have a feeling this algorithm will get better, but a 20% premium is still a small price to pay for insurance that some lawyer won't come knocking on your door, ever, demanding royalties.

--

Alot's happened since then

[xiphmont]

If you're not keeping up, please don't quote obsolete facts as current truth.

The Slashdot posting in April was prealpha code. Substantial development and tuning has happened since then; not only is average Ogg filesize now *smaller* than mp3, the audio quality is much improved. We're now four days from 1.0 beta. Go get it and see for yourself if you don't believe me.

Monty

Why I should use Vorbis!

[MicroBerto]

1. Open source
2. No royalties ever owed to jerks
3. Already comparable to mp3 and once finalized will be better quality and compression
4. LGPL
5. It will support more than 2 channels (if it doesn't yet)
6. Already have audio player plugins for it, they're getting better.
7. With a name like Ogg Vorbis, its GOT to be good!
8. Their logo is awesome

and finally..

9. The extention is going to be .ogg !!! How cool is that?! I'm there!

Mike Roberto (roberto@soul.apk.net) -GAIM: MicroBerto

The sweet bliss of new standards...

[DranoK]

The coolest thing about this isn't about the free as in beer format, but the fact that it's an open format. I think more and more compression/encryption/etc in the future is going to abide by these open formats, as png has led the way (and hopefully this will continue the cause).

My first thought, of course, was that no new standard could compete with mp3. Then I realized it wouldn't really have to. Let's look at gif vs png, for example. Average Joe User coasts through webpages, never ever knowing if the pretty images he sees are gifs or pngs. He doesn't need to; all he cares about is that the pictures are there. Netscape and IE are the ones that had to do the work of getting png support for their browsers else worry about being slammed as "incomplete". The same will happen with any new open standard, I think.

Winamp, for instance, isn't going to stop playing mp3's. But I bet a future version of winamp will support the new format as well. This is the power of open standards.

Format types are a pain in the ass when they are closed. No company wants to buy liscencing rights to add support for a format, and will only do so if the format is so huge that their product won't sell without it. This stalls development of free software (if you're not going to be paid for it in the end IMHO the less likely you are to shell out thousands of bucks for a liscence) as well as a stagnation in standards. If all standards were proprietary, creating a new one would be hard to get out since companies would rather only impliment those standards that are set in stone. They likely won't purchase a liscence to a new commercial standard before it's been proven, and it won't be proven until it's liscenced. Without open standards stagnation would prevail.

For one last example look at the competition between OpenGL and GLIDE. GLIDE was too proprietory, and after a few years of fame slowly slipped away into the night. The industry is realizing this now, and is embracing open standards.

If only they'd realize the same logic applies to Open Source as well...

DranoK
Having honestly nothing better to do today than read old Slashdot stories ;)



That is not dead which can eternal lie, and with strange eons even death may die.

iCast is helping out on legal

[xiphmont]

You're not the first to 'call' on this issue :-) And you definately should.

The legal review of Ogg and Vorbis patent status is one of the things iCast is helping with. I don't know how much of the patent review will be on the website at 1.0 beta-time. For the most part, I've been keeping my head technical and not tracking the publicity or legal push going on around me at iCast; I know that the lawyers so far are very comfortable with the patent review, but I don't know what documents they've produced so far to prove we're not just bluffing :-)

Of course, results of the legal review will be public knowledge as soon as it's finished. So far, no surprises (I have more patent summaries in my inbox to review right now...)

Monty

Re:"CD" quality? Not.

[Pope]

CD quality was never the goal of MP3: 10:1 compression of a digital source was.
Personally, i can't stand the goofballs who post 256 and higher MP3's on USENET. It doesn't significantly change the quality enough to merit the bigger files.

Simply having another music sound encoder that's open source will not solve the higher problem of the massive copyright violations that MP3 brought about.

Having said that, I will say that I do post and download lots of MP3s on a weekly basis, I just don't try to justify it with weak excuses and "reasons." I do it simply because I can. If I didn't have DSL, I wouldn't bother. Hell, I had ADSL for almost a year before I even thought of looking for them on USENET. I mean, isn't that where all the pr0n is? :)

Pope

Freedom is Slavery! Ignorance is Strength! Monopolies offer Choice!

Jack the Ogg pimp

[xiphmont]

Jack Moffitt is overseeing Ogg's production and implementation within iCast. He's one of those truly rare breed of geek; he not only has a technical whip (and knows how to use it), he enjoys talking to people, drumming up support, evangelizing and remembering all the names at a meeting.

Ogg got written because of me... but Ogg is getting the attention it deserves because of Jack.

Monty

Cheap Karma

[Andy_R]

Well, if this lot got a (5, informative) and two (4, insightful)s last time, it should be worth a try....

"Here's a link to an interview with the author, with his explanation of why vorbis is better than mp3. http://www.advogato.org/article/56.html

One thing that everyone seems to be missing, is that Vorbis supports bitrates of 16kbps-128kbps per channel! Since it uses better algorithms than MPEG-1 Layer 3 (MP3), it has the potential to sound much better. It's not done yet, and the development team is still making changes to it that will affect the quality. I'm going to wait and see how it works, but it sounds like it will be excellent when it gets done.

It's terrific to finally see an open, IP-free audio codec with (seemingly) great sound and compression efficiency. One of the things most often complained about at Slashdot is the lack of Quicktime players for Linux, and more specifically, lack of a player capable of playing moviescompressed with Quicktime 4's Sorenson codec. Many sites, especially those of the movie industry, have adopted Sorenson because it has genuine advantages over industry-standard MPEG video: Sorenson produces significantly better video quality at the bitrates preferred on the Internet today. While Sorenson and Microsoft's proprietary offerings are gaining ground, the use of free video standards like MPEG is becoming more and more scarce.

The only feasible way of reversing this trend is to come up with a superior video codec and distribute it freely. Until now, many people have argued that developing a good media codec involves such high-end mathematics that developing one under traditional Open Source development model is not possible. It is high time that someone proved them wrong."

...oh all right then, I'll settle for (score:13, reundant)

- Andy R.