Instead Of Recording Music, Recreate It 1,000 Times More Efficiently Than MP3s (But Only For Clarinets)

from the that's-gotta-lose-some-quality dept

Thu, Apr 3rd 2008 2:21pm — Mike Masnick

There's a report going around about some researchers who created a music file that's apparently 1,000 times smaller than MP3s. However, what's really interesting is how it's done. Rather than actually record the sound, it's designed to recreate the sound itself. As the article notes, it's more like the way a player piano plays music from a roll of punched paper than a recording of the original piano. Of course, that's got to make you wonder about the quality, and whether it comes out sounding mechanical, losing the actual nuances of how the music was played -- but the researchers insist their system captures exactly how the music was played as well. Don't expect to hear much out of this research for a while, though. Right now, the system is rather limited. It can only work with certain types of music (clarinet music from the sound of it), because they had to program in the specifics of how a clarinet is played (such as fingering, breath pressure, and lip pressure). Considering that they would then need to do that with every single instrument, somehow this doesn't seem likely to be in practical use any time soon.

Thank you for reading this Techdirt post. With so many things competing for everyone’s attention these days, we really appreciate you giving us your time. We work hard every day to put quality content out there for our community.

Techdirt is one of the few remaining truly independent media outlets. We do not have a giant corporation behind us, and we rely heavily on our community to support us, in an age when advertisers are increasingly uninterested in sponsoring small, independent sites — especially a site like ours that is unwilling to pull punches in its reporting and analysis.

While other websites have resorted to paywalls, registration requirements, and increasingly annoying/intrusive advertising, we have always kept Techdirt open and available to anyone. But in order to continue doing so, we need your support. We offer a variety of ways for our readers to support us, from direct donations to special subscriptions and cool merchandise — and every little bit helps. Thank you.

–The Techdirt Team

Filed Under: clarinets, mp3, recording

34 Comments

If you liked this post, you may also be interested in...

Reader Comments

Subscribe: RSS

View by: Time | Thread

Captain Obvious, 3 Apr 2008 @ 2:50pm

Wow, if only someone had invented this 25 years ago and called it MIDI. Oh wait.....
[ link to this | view in chronology ]
- Mike (profile), 3 Apr 2008 @ 3:48pm
  
  Re:
  Wow, if only someone had invented this 25 years ago and called it MIDI. Oh wait.....
  
  MIDI is more about synthesizing. This seems to be more about being able to match what's actually played...
  [ link to this | view in chronology ]
  - Anonymous Coward, 3 Apr 2008 @ 10:12pm
    
    Re: Re:
    MIDI is more about synthesizing. This seems to be more about being able to match what's actually played...
    
    MIDI is also about what's actually played. In fact, there are MIDI recorders that record what's played by musicians playing MIDI instruments.
    [ link to this | view in chronology ]
Lord, 3 Apr 2008 @ 2:58pm

Now try doing it with voice
[ link to this | view in chronology ]
Casey, 3 Apr 2008 @ 3:03pm

I'm a Mac
Garageband.
[ link to this | view in chronology ]
skyrider (profile), 3 Apr 2008 @ 3:10pm

21 years ago, give or take
and it's called MOD

http://en.wikipedia.org/wiki/MOD_(file_format)

a little bigger than midi, because it carries the sound samples with it.
[ link to this | view in chronology ]
Free iPhone, 3 Apr 2008 @ 3:10pm

Please explain....
I have read the article three times already, what exactly is it trying to say !
[ link to this | view in chronology ]
Robert, 3 Apr 2008 @ 3:11pm

Well
They are right, this sounds exactly like the MIDI standard that's been around forever.

The Midi file contains the notes and timings that need to be played and the playback software applies that information to an instrument (Software or hardware) and recreates the sound.

What's cool is that you can put tracks on instruments they were not intended for like playing the piano track on a guitar for example.
[ link to this | view in chronology ]
Adam Singer, 3 Apr 2008 @ 3:21pm

MIDI
http://en.wikipedia.org/wiki/MIDI

yup - i just rendered some midi out of my studio

its a whole song worth

its 5K...im not a researcher. it has all the notes and the attack on each note
[ link to this | view in chronology ]
- Steve, 3 Apr 2008 @ 5:01pm
  
  Re: MIDI
  Awesome - A new way for the unknown musician to make money off of their songs.. Instead of publishing them, get a "research grant" and publish the .MID file as your product...
  [ link to this | view in chronology ]
bupbe, 3 Apr 2008 @ 3:24pm

I got it!!!!!
The whole scheme is called "Cpoy"
[ link to this | view in chronology ]
Jerkometer, 3 Apr 2008 @ 3:30pm

When was this posted
Looks like an April Fool's joke if you ask me. 1990 called and wants it's breakthrough back...ugh that was a horrific joke.
[ link to this | view in chronology ]
- Ima Fish, 4 Apr 2008 @ 5:39am
  
  Re: When was this posted
  It is an april fool's joke. I wonder whether Mike knew that.
  [ link to this | view in chronology ]
  - Anonymous Coward, 4 Apr 2008 @ 3:48pm
    
    Re: Re: When was this posted
    It is an april fool's joke. I wonder whether Mike knew that.
    
    Considering that Mike posted this on the 3rd, he can't claim only going along with the joke. Even funnier than the original joke was watching Mike try to defend his writeup and pretend it this was something different from MIDI. He really has a hard time admitting when he's wrong. I suppose that would undercut confidence in the "analysis" he sells.
    [ link to this | view in chronology ]
James Yu, 3 Apr 2008 @ 3:38pm

Been Around for Years, but it's not MIDI
It's not MIDI because their research is more about the physical modeling of an instrument. But this type of synthesis research has been around for years. Probably the only reason this gets an article is because they're using much more sophisticated modeling than any off the shelf synth (at least for clarinets) that you can get.

Still though, nothing earth shattering. You can get a more general algorithm by using something like additive synthesis: http://en.wikipedia.org/wiki/Additive_synthesis

But it probably won't sound as good for their clarinet case.

Who wants to only listen to clarinet's anyway?
[ link to this | view in chronology ]
Boris Jacobsen, 3 Apr 2008 @ 3:52pm

Two-way algorithm....
When do we see the algorithm that takes a piece of recorded music and converts it completely accurately to even just clarinets. Add a few more instruments (a full orchestra, say) and some unique human voices and, if this method is even possible, which I VERY MUCH doubt (how do you model every known instrument and every human voice?), then the files are going to be a lot larger than an mp3 file.
[ link to this | view in chronology ]
- Nasch, 4 Apr 2008 @ 1:28pm
  
  Re: Two-way algorithm....
  Presumably you would install the library of instruments once, so it could be huge. Then the very small MIDI-like file would instruct the player program which instruments to use and how. What I'm wondering is who's going to pay for the research needed to reproduce the sounds of (at least) dozens of kinds of instruments? After that work is done it seems like this could go forward effectively.
  
  Unless it's an April Fools joke like someone claimed. I have not, of course, RTFA. Hope that acronym works on TechDirt the same as /.
  [ link to this | view in chronology ]
Jonatas Miguel, 3 Apr 2008 @ 4:31pm

This isn't exactly like MIDI, note that in the article the writer clearly states that they recreated the fingering and breathing techniques, which is something you cannot do with MIDI, or at least, you can't get a realistic sounding sample anyways... Have you ever heard a string ensemble in MIDI, it doesn't sound correct, the movements between notes aren't fluid like they are when played by a professional violinist or such, and basically that's what these researchers are trying to recreate... It's my way of seeing it anyways.
[ link to this | view in chronology ]
- Anonymous Coward, 3 Apr 2008 @ 10:17pm
  
  Re:
  This isn't exactly like MIDI, note that in the article the writer clearly states that they recreated the fingering and breathing techniques, which is something you cannot do with MIDI, or at least, you can't get a realistic sounding sample anyways... It's the same basic idea. And even basic MIDI includes stuff like "velocity" for stuff like fingering. If you need to get more complex you can layer multiple channels to get additional intonation effects.
  [ link to this | view in chronology ]
Anonymous Coward, 3 Apr 2008 @ 4:57pm

Mellotrons
Haven't Mellotrons been doing this for years?
[ link to this | view in chronology ]
matt, 3 Apr 2008 @ 5:12pm

not even a remote invention
1000 times smaller than an mp3? at what bitrate? 5000kbps?

At some point the sampling rate comes into play here. This company wants to pump and dump probably. If only they'd come on techdirt, flame us all and defend how "impressive" their innovation is so we could get a good read :D
[ link to this | view in chronology ]
zcat, 3 Apr 2008 @ 8:29pm

still sounds like midi to me
OK, so they added a few extra parameters. The basic concept is still more or less identical to the concept behind midi. Really old technology re-implemented with a few obvious improvements is still really old technology.
[ link to this | view in chronology ]
Anonymous Coward, 4 Apr 2008 @ 12:46am

MIDI Clarinet
A friend of mine has a MIDI keyboard that he can load samples of almost any kind of instrument into, including clarinets.
[ link to this | view in chronology ]
Jake, 4 Apr 2008 @ 2:28am

Even if it is just a massively upgraded version of MIDI, it still sounds pretty impressive on paper. I'm reserving judgement until I've played one of those files through my own soundcard and speakers, though.
[ link to this | view in chronology ]
- Anonymous Coward, 5 Apr 2008 @ 10:03pm
  
  Re:
  Even if it is just a massively upgraded version of MIDI, it still sounds pretty impressive on paper.
  
  Massively upgraded? It pretty much described MIDI as it is now. What a joke. In fact, it was an April fool's joke. Seems there are a lot of fools about.
  [ link to this | view in chronology ]
  - Mike (profile), 6 Apr 2008 @ 9:59pm
    
    Re: Re:
    
    Massively upgraded? It pretty much described MIDI as it is now. What a joke. In fact, it was an April fool's joke. Seems there are a lot of fools about.
    
    Do you have any proof it's an April Fool's joke? I've been looking for some, and haven't found any. In fact, Wired confirmed with the University that it is not an April Fool's joke:
    
    http://blog.wired.com/music/2008/04/new-codec-crams.html
    
    If it is, I'll change the post, but so far I've seen nothing to indicate that this is false, and a few sources that have indicated that this is different than MIDI.
    [ link to this | view in chronology ]
Rick Sarvas, 4 Apr 2008 @ 7:44am

More compact format than MP3
I remember hearing about something a few years ago where someone came up with a way to use a series of vector Bézier curves (http://en.wikipedia.org/wiki/B%C3%A9zier_curve) to trace the content of a sampled wave file. Just imagine it as tracing the magnified view of what you see on the screen in an audio editing program like SoundeForge with a vector drawing program like Illustrator or Inkscape - but do it for the entire length of the file. In theory, you could reduce a given number of samples that make up part of a sine wave down to just a few data and control points. The resulting output wouldn't have anywhere near the fidelity of the original audio sample and would probably not be suitable for music, but it might be a good way to compress other types of audio where quality is not so important, say for example, spoken text.

I wonder how well this might work for compressing analog video signals.
[ link to this | view in chronology ]
Iron Chef, 4 Apr 2008 @ 4:29pm

Interesting.
I've watched this post since it first appeared.

A 20-second clarinet sound in 1 kilobyte sounds more like a "recipie" than the actual waveform.

If that's the case, there's probably resource-intensive additional hardware or software that interprets the "recipie" into a waveform thru the way of "synthesis".

The hardware, is probably called a "voice table" if it's on a soundcard, or "synthesizer" if it's standalone.

Chances are, the hardware looks something similar to this:
http://www.bhphotovideo.com/c/product/372560-REG/Yamaha_MOTIF_RACK_ES_MOTIF_RACK_ES_.html

:-) To #25: at least Mike doesn't delete his mistakes and shows us what it's like to be human.
[ link to this | view in chronology ]
- Anonymous Coward, 5 Apr 2008 @ 10:21pm
  
  Re: Interesting.
  :-) To #25: at least Mike doesn't delete his mistakes and shows us what it's like to be human.
  Notice that Mike has yet to admit his foolishness. And to me, that just makes it even funnier. :)
  
  He doesn't delete his mistakes because he's seen what the "Streisand effect", as he has called it, can do. In other words, it would only make it more well known, which is something that I doubt he wants. :)
  
  But you're right about being human. It's sometimes hard to admit that you're wrong, as Mike is aptly demonstrating. :)
  [ link to this | view in chronology ]
  - Mike (profile), 6 Apr 2008 @ 10:01pm
    
    Re: Re: Interesting.
    Notice that Mike has yet to admit his foolishness. And to me, that just makes it even funnier. :)
    
    As I said, I'm waiting for proof that this is an April Fool's joke. So far, I've seen multiple reporters who have confirmed this is real. If it's actually an April Fool's joke, that would be quite interesting, and I'd even be willing to post a follow up to call more attention to it, because it's interesting (even if I look foolish for it...). It's just that no one has shown any proof that it is an April Fool's joke.
    [ link to this | view in chronology ]
    - Anonymous Coward, 7 Apr 2008 @ 12:48am
      
      Re: Re: Re: Interesting.
      So far, I've seen multiple reporters who have confirmed this is real.
      
      It's real all right. Just not new.
      
      It reminds me of an article one of the electronics trade magazines (IEEE?) published one April 1st many years ago back when laser printers were new and very expensive on how to make your own for less. It went into great detail about how to remove the lamp from an old photocopier and then attach a CRT computer monitor face down to it so that the image from the monitor was supposedly then picked up and printed out by the copier. Like I said, it went into great detail about how to get the focus right, adjust the exposure times, synchronize the CRT scan rate, compensate for the curvature of the CRT, etc.. The seriousness and detail made it really look genuine. I was in a university electrical engineering program and someone posted the article on the wall in the senior design lab. It was really funny watching students who should have known better gathering around and trying to figure out where to find an old photocopier. :-)
      [ link to this | view in chronology ]
      - Anonymous Coward, 7 Apr 2008 @ 1:24am
        
        Re: Re: Re: Re: Interesting.
        To add to my own post (sorry)...
        
        The thing about jokes like this is that "proof" that it was a joke often never comes. I wouldn't be surprised if some of the folks fooled by the DIY laser printer story still believe to this day that it was true. I wish you'd identify those "multiple reporters who have confirmed this is real." And by real I mean "for real that it's new". I'd like the opportunity for a few more laughs. :-)
        [ link to this | view in chronology ]
LYCANTROPUZ, 6 Apr 2008 @ 4:26am

1000 TIMES SMALLER?......MORE ROM TO PUKE ON....
HI....
INTERESTING ARTICLE....
I JUST THINKIN' IF SOMEONE BRUTALY-COMPRESS MUSIC NO MATTER HOW IT IS DONE... THE FACT ABOUT MASS-COMPRESSION OF MUSIC (OR FILES, OR "RAW DATA") IS THAT YOU'LL GET SMALLER FILE, AND OF COURSE, MORE ROM IN YOUR DATA-STORAGE-DEVICE (MP3 PLAYER, IPOD, GP2X, MOCO, ETC..)
JUST THINK ABOUT IT: YOU HAVE A BEEFFY 8 GIGS PM3 PLAYER AND SMASH-POTATO-COMPRESSED' MUSIC FILES, YOU MUST EXPEND NEAR 8000 $$$ PURCHASING MUSIC (99 LINCOLNS EACH)ALSO MUST EAR FOR 23 DAYS & NIGHTS ALL MUSIC WHITOUT REPEAT ANY AND YOUR MP3 PLAYER GETS BARELY HALF DONE BELIEVE ME. MP3 BECOME MP33 AND STILL HAVE ENOUGH ROOM (4GIG) TO PUKE IN, I MEAN, TO PUT MANY "MEDIA" EXTRA OR ONE(1) OPRAH SHOW(4.7GIG)(SAME TO WASTE RESOURCES).
NOW, THE POINT IS:IF PORTABLE MEDIA IS GOIN' FAT ("GIGAS" OF STORAGE), THEN, WHAT IS THE POINT ON MASS-TIGHTEN MUSIC FILES?...
THANK YOU,, HUGS FROM CHI-IL
P.S. FORGIVE MY HORTOGRAFIC ERRORS PLS
[ link to this | view in chronology ]
- LYCANTROPUZ, 6 Apr 2008 @ 4:38am
  
  Re: 1000 TIMES SMALLER?......MORE ROM TO PUKE ON....
  IT'S ME AGAIN....
  HAVE U EVER SEEN "ANIMUSIC" AND READ HOW ITS DONE?
  IT USES AN ALGORITHM TO PLAY MUSIC FROM MIDI TRACKS, SAME AS "CILINDRO" (KIND OF MUSICAL CILINDER WITH SMALL SCREWS IN ITS SURFACE) SOMETHING LIKE BRAILLE FOR BLIND PEOPLE.
  AND IT A GOOD SANPLE FOR MAKING ULTRA-COMPRESSED MUSIC
  [ link to this | view in chronology ]