Researchers have digitally reproduced music in a file nearly 1,000 times smaller than a regular MP3 file.

What will they think of next?

They've got a ways to go before they start to put musicians out of business but if this technology succeeds to the point of not being able to differentiate between real music and its virtual counterpart it could revolutionize the music and recording industry as we know it. And I don't mean in a good way. What do you think? I know it sounds a little robotic at present but if they perfect it look out.

Best,
Mike



[This message has been edited by keybplayer (edited 04-02-2008).]

We're doomed anyway & been on a path of destruction son the mid 70's....so bring it on maybe it will bring a climax & reverse itself back to normalcy.

In other words it’s a sound modelled Clarinet instead of a sampled one, so what’s new? Sound modelling has been around for a few years now; quite a few VSTs use it, as does Wersi with its OAS 7 software. (This is the reason that the new voices only take up a fraction of memory compared to the previous sampled sounds)
Try True Pianos http://www.truepianos.com/ which produces a sound similar to a 2 or 3GB sampled piano with a fraction of the memory.
Another one is the Miles Trumpet http://www.soundfonts.it/?a=read&b=25
or http://www.sonic-core.net/en/products/soniccore.html

Bill


[This message has been edited by abacus (edited 04-02-2008).]

Reckon that's what will happen. People always like to be entertained by other people for the most part in my view, not just technology. It's the combination of a person AND the technology that makes the entertainment, so I think people will always be involved.
How we, as entertainers, choose to adopt and modify what we do and how we do it is totally up to us.

Remember the old saying
Technology is a very good servant
But a very bad master

Bill

Within five years you will be able to fit all of the music every produced and all of the movies ever produced on what harddrive that fits in your pocket. We don't need to wait for this new technology for a revolution in the music industry.

The file to make the sound may only be 1kb, but the application to replay it is probably Megabytes (and I'm not sure just how many clarinet solos you are going to want to walk around with!).

This is like saying that a MIDI file can do an entire piano concerto in a few kb's, while ignoring the gigabytes of data needed to play it...

BTW, anyone notice the date of this 'announcement'?

Well spotted, I missed it

Bill

[This message has been edited by abacus (edited 04-03-2008).]

I think you are mis-interpreting the technology used in making the music. The audio recording is not gigabytes of data needed to play it because it is not real audio data that is used in the first place, at least not in the file that is compressed anyway. It is not the audio itself in the original real clarinet recording used to make the compressed file but rather using a computer model a computer literally reproduces the original performance based on everything it knows about clarinets and clarinet playing i.e. the physics of a clarinet and the physics of a clarinet player and then reproduces it accordingly by way of a virtual model. Once this technology is spot on in its reproduction capabilities for any given instrument known and used in todays world and even those from yesteryear the real instrument would then be unnecessary in the recording of subsequent audio because the virtual instrument will have "learned" to reproduce the sound just as well as the real thing - without the tons of audio data necessary to record actual real instruments in an audio environment.

As far as putting musicians out of business, if this technology succeeds in a big way session players could become obsolete in my opinion, as well as musicians used in other recording venues such as in various types of Studio recordings. In other words, once the virtual instrument "learns" the behavior from its counterparts i.e. real instruments, thereafter the real instruments wouldn't be needed at all to produce music for the virtual modeling counterpart. If the virtual instrument(s) know(s) everything about the real instrument(s) already, both in the physics aspect of the instrument itself and the physics of the instrument player used to play the instrument(s), then only the virtual counterpart would be necessary to reproduce the music and/or sound of any given instrument, without the need of using real instruments at all to do so. At least so the theory goes anyway.

PS: I choose not to dabble in superstition. Neither does Stevie Wonder by the way. Next thing you know there will be conspiracy theorists coming out of the woodwork too.

Best,
Mike

[This message has been edited by keybplayer (edited 04-04-2008).]

Have you USED any modeling software?

It's a tradeoff... You don't need gigabytes of sample data, but you need vast amounts of computing power. The more accurate the model, the more power you need. And they still have a VERY long way to go to come out with ANY acoustic instrument model that will fool the listener (you didn't think for one minute that sounded like a REAL clarinet, did you?).

Let's take the piano, shall we... Pianoteq by Modartt http://www.pianoteq.com/ is a modeled piano. It's pretty good, and actually does a FEW things better than samples (string resonance, for instance), but compare it to Ivory http://www.ilio.com/synthogy/ivory/ and you realize it still has a way to go.

But the assertion that we were listening to MUSIC was absurd. We were listening to a very poor modeled clarinet. For some, that may indeed BE music. I would think, though, if you wanted to listen to music, you might want a few more instruments...! This is where the article blue-sky's us. The horsepower to even model ONE instrument truly accurately doesn't yet exist. You can get close, in a few cases, but no banana. But the horsepower to model an orchestra, even a small wind ensemble is decades away.

In the meantime, major Hollywood movies are being scored by sampled orchestras. You are being fooled by them every day on TV. This is NOW.

Modeled orchestras are 22nd century technology...

No, I didn't think it sounded like a real clarinet Diki as I stated in my original post i.e. I thought it sounded robotic. I agree that they have a ways to go but as far as taking another 100 years I'm not so sure. Theoretically computing power doubles roughly every 18 months or so (Moore's Law) but realistically Moore's Law has technically reached a threshold and until new discoveries in ways to increase processor power are attained Moore's Law may become an obsolete gauge to rely on in the meantime. But eventually those barriers will be overcome and I don't think it will take 100 years but more in the neighborhood of 10 to 20 years at most and we may be surprised by such advancements in as little as 5 or less years.

>> Yeah, I know sampled music is used everywhere e.g. movies, cd's, etc. but the music that is sampled is done so from real instruments. Which is a good thing since I am all for keeping real musicians playing and using real instruments. But if the technology gets to the point where real instruments are no longer needed in a multitude of various recording and live venue applications then of course real musicians will lose out unfortunately.

It's like cotton pickers. They were needed until the cotton gin was invented. I guess they call it progress. Regarding this technology for making music, I call it 'regression' if you ask me. At least in how it pertains to musicians anyway.

Best,
Mike

If you read my posts you will find sound modelling has been around some time and with the increase in processing power is starting to increase rapidly. (Wersi get round the processing problem by using a dedicated card and not relying on the CPU to perform the calculations)
10 years from now I would be extremely surprised if sampled instruments were still around in even entry level boards. (The more expensive boards will still have better sound modelling however)
As to replacing musicians, not a chance, the musician will be able to express himself a lot more, and in a way that even the best samples cannot reproduce.

Bill

One can only hope...

Existing successes in the modeling software are mostly confined to modeling synthesizers and tone-wheel electro mechanical keyboards, and some tine-based pianos (Rhodes and Wurli's, primarily), which have VASTLY simpler behavior properties than acoustic instruments.

Can anyone point me to ANY acoustic instrument model that is truly capable of fooling the listener? At a wide range of musics and performance techniques? I can't honestly say I've heard one yet...

Moore's Law is starting to bump into theoretical limits, so you have to posit an entirely revolutionary technology that no-one has yet successfully developed to go with the idea of even ONE instrument being emulated successfully. One clarinet... one super computer. Two clarinets, two supercomputers (each model is totally independent, two voices takes twice the horsepower)...

You can see how far things are going to have to go to emulate an orchestra. I'm sorry, but I don't see this in my lifetime.

But i guess the average symphonic clarinet player is not the slightest bit concerned with whether his job is replaced by modeling, or by sampling. He's out of a job, either way...

Hi Diki
Technology and computers is my day job, and I can assure you that within 10 years, sound modelled sounds will be the norm. (You will literally have the power of the old Cray Super Computers in your mobile phone)
As to getting rid of musicians, then they have been saying this for decades, (Whenever new technology comes out) however in practice good musicians just embrace the new technology and use it to their advantage, thus forcing the R & D departments to bring out further technological improvements. (It’s just a standard cycle)
Regards

Bill

Perhaps you aren't aware of the hardships that orchestral musicians face nowadays, compared to thirty years ago.

Every single film and TV show that featured orchestral music used the real thing (there were plenty of synth and dance pop scores, even then, though) back then. This is NOT the case, right now, and few orchestral musicians can support themselves (compared to the 'golden days') by playing alone.

This isn't the complete demise of the system, but it IS a substantial collapse. Ask any 60 year old clarinetist living in LA...!

The thing that has always struck me about modeling is that, the more accurately you want to model the sound, the more input parameters need to be used. A real acoustic instrument is a very chaotic system, with vast differences made to the sound from a plethora of different sources. Take your clarinet (by FAR one of the simplest 'tones' to emulate);...

To emulate a clarinet PERFORMANCE, rather than just the clarinet 'sound' (sampling can do that easily enough!), you have to model just about every performance technique the clarinet player uses. Tonguing, reed pressure, wind support, 'bite', reed angle, fingering technique (clean, slid, fluttered, etc.), a myriad of other ways to influence the sound, and how the notes are joined up and produced.

For the model to have the ultimate realism, EACH of these variables needs to be controlled independently, and what performance tool offers so much simultaneous controls...? Certainly not any keyboard. Even wind controllers are primitive in their ability to send codes for every one of these parameters. No... the best controller to be able to control a clarinet emulation model is... a clarinet!

Kind of defeats the point, doesn't it?

Not only do you have to posit a computing technology that is the stuff of sci-fi to be able to emulate more than just ONE instrument, and a modeling technique that is light years past the point where we are now (still haven't been pointed to ONE accurate model, yet), you also have to posit a controller for the model that is virtually as complicated to play as the real thing...

And, once again, with current technologies, you can see the huge leaps ahead in sampling that are being used by the major orchestral sampling libraries (and in a small way, the SA voices on the T2/S900) by switching between different samples of different performance techniques in real time. THIS 'state of the art' is already capable of fooling the listener. No need to develop something that would be MUCH harder to control, MUCH more processor intensive and has still to be determined if it CAN go the whole way (or even as far as sampling has already gone)....

But ask that LA clarinetist what he's most afraid of... modeling, or samples? I'm pretty sure I know what he'd tell you...

Hi Diki
I agree the amount of work has reduced, but it still flourishes well on this side of the pond. (Different states of mind I suppose)
Regarding modelling, then yes it is difficult to work out the mathematical algorithms, but musicians and engineers are slowly starting to get the hang of it.
Found this http://www.samplemodeling.com/en/demos.php while I was doing a bit of surfing, which combines both sampling and modelling.
Certainly looks interesting
Regards

Bill