 The Clarinet BBoard
|
Author: skygardener
Date: 2008-04-03 14:22
I thought this was going to be one of those "the day will come when musicians are no longer needed" things, but after hearing it all I can do is laugh. It sounds like a synth from 15 years ago!
|
|
Reply To Message
|
|
Author: Tobin
Date: 2008-04-03 14:34
...I wonder what kind of mouthpiece the midi is using???
James
Gnothi Seauton
|
|
Reply To Message
|
|
Author: crnichols
Date: 2008-04-04 16:48
I thought we were supposed to be trying to make things sound better, not worse... And isn't technology supposed to produce "perfect" intonation. A few of those notes were way out of tune.
Christopher Nichols, D.M.A.
Assistant Professor of Clarinet
University of Delaware
|
|
Reply To Message
|
|
Author: Paul Aviles
Date: 2008-04-04 18:51
Even if we don't just disparage the end result on specifics, don't all the gagigabytes of ROM/RAM needed to interpret the clip count against their supposed savings of file space?
I hate computers.
...............Paul Aviles
|
|
Reply To Message
|
|
Author: tictactux ★2017
Date: 2008-04-04 18:58
Paul Aviles wrote:
> Even if we don't just disparage the end result on specifics,
> don't all the gagigabytes of ROM/RAM needed to interpret the
> clip count against their supposed savings of file space?
That "compressed file" is not much more than the notes on the sheet...every reduction in instructions is bought with additional complexity in interpretation.
I don't really see the benefits of all this research, and I'm a geek, somehow.
--
Ben
|
|
Reply To Message
|
|
Author: Copland
Date: 2008-04-04 19:49
Yeah, agree with tictactux.
Storage rates have expanded so much that it is not terribly important to have files that are that small. Plus, its really the playback of the data that's most important, and for a file that small you've got to imagine it takes a LOT of processing power to interpret such a sparse amount of data.
Plus, my take on it is that when you buy a CD (or record, or mp3, or cassette tape, or whatever), it's almost always because of who's playing it. I don't want a computer to pretend to be playing like an artist I like, I want an actual recording of the person.
|
|
Reply To Message
|
|
Author: Phat Cat
Date: 2008-04-05 10:53
Well, I have the ultimate compression technique: it reduces any file to a single bit. I'm still working on the decompression algorithm.
Seems like the same problem in the article.
|
|
Reply To Message
|
|
Author: monzamess
Date: 2008-04-05 18:41
As someone who's been interested in computer-generated music since I was a kid, I tried to approach this report with an open mind. Maybe the article misses the main point because this is really underwhelming. I agree that reducing the file size is not very important (neat as it is) and can't really be the main goal of their research.
I think the main point is, "Then, says Bocko, it's a matter of letting the computer 'listen' to a real clarinet performance to infer and record the various actions required to create a specific sound. The original sound is then reproduced by feeding the record of the player's actions back into the computer model."
On the one hand, this is neat because they have a model that "learns" how to play music. On the other hand, it's not such a great approach because they would need to model every instrument they wish to reproduce, and so far, they don't seem to be doing too well with just the clarinet. That sample clip wasn't anywhere close to listenable--this coming from someone who listened to hours of 3-voice-plus-noise-generator-square-wave music on my PCjr way back when, and liked it!
|
|
Reply To Message
|
|
Author: DavidBlumberg
Date: 2008-04-05 19:17
I was doing electronic music in 1977. What I heard wasn't much better... Piano can be reproduced quite well, but not Clarinet even with the Garritan synths. Which are close sortof, but can't fool a Clarinetist.
Hey, there's a clip of a computer playing Giant steps on Sax on Utube. The computer is actually playing it moving the keys and blowing air into the Mouthpiece. I'm writting from my phone son can't link it, but a search would find it.
I found that pretty cool. If I recall the original link was posted here by John Moses so you could search Giant steps on the board search to get the link.
http://www.SkypeClarinetLessons.com
|
|
Reply To Message
|
|
Author: mrn
Date: 2008-04-07 21:08
One of the problems with this article is that they are referring to this technology as "compression." It's not really compression. What these folks have apparently developed is more like a kind of "notation" for things like lip pressure, air pressure, etc. In that respect, it's more like MIDI, except much more detailed. Comparing this to an MP3 is like comparing sheet music to a CD. A far better comparison would be to a player piano roll.
The really "gee whiz" part of this technology, though, is what monzamess mentioned--the computer listens to the human performance, infers what precise physical actions were needed on the part of the human performer to produce the given sound, then records the physical actions.
Unlike the folks who wrote the article, I tend to think that the real value of this kind of research is not in sound synthesis, but in being able to accurately simulate a musical instrument for the purpose of understanding how the instrument works and for making improvements in the instruments themselves and in technique. It's much easier to make changes to a computer simulation of a clarinet than to a prototype. Likewise, if you had a detailed realistic model of a clarinet and player, you could do things like do computerized searches for alternate fingerings or multiphonic fingerings, etc.
|
|
Reply To Message
|
|
Author: C2thew
Date: 2008-04-07 22:18
http://www.youtube.com/watch?v=OjONQNUU8Fg
giant steps played by a robot.
How do you reproduce a file that small while still maintaining the quality? i really can't see anyone making a comeback with midi files.
Our inventions are wont to be pretty toys, which distract our attention from serious things. they are but improved means to an unimproved end, an end which was already but too easy to arrive as railroads lead to Boston to New York
-Walden; Henry Thoreau
|
|
Reply To Message
|
|
Author: Ken Shaw ★2017
Date: 2008-04-09 11:34
Um... er... uh... The article is dated April 1, you know...........
|
|
Reply To Message
|
|
Author: Steve Epstein
Date: 2008-04-10 05:30
Ken Shaw wrote:
> Um... er... uh... The article is dated April 1, you
> know...........
Yeah, but it still plays:)
Steve Epstein
|
|
Reply To Message
|
|
Author: crazy karlos
Date: 2008-04-10 08:47
Oh, this was quite serious, I'm afraid -- it was first reported in the J Acous Soc Am in November 2007 (http://lib.bioinfo.pl/pmid:18190015).
I also think the notion that this is actually about "compression" is very misleading. I am honestly not sure what these people think they're doing. The only *real* smart thing they did -- given that this is in fact a serious project -- is start with the clarinet.
I mean, if they're actually serious about "compressing" all kinds of recordings in this manner, they could have started with the human voice -- and where would *that* have taken them?
I couldn't listen to this "performance" (no streaming media allowed at work ... ) but the fact that the intonation is wonky, shows they may be closer to true clarinet physics than they realise ...
What occurs to me is this: the "Expert System". Let's suppose that they take some great clarinetist, and integrate absolutely everything that this person ever recorded into the programme, so you have the complete "virtual performer" -- let's say Artie Shaw, who maintained "I did all you can do with a clarinet. Any more would have been less."
Then if a new composition appears, you give it to the machine, and get the definitive Artie Shaw interpretation.
If this programme were available to you, you could make sure that all your very best characteristics were "uploaded" -- let's say you find that once-in-a-lifetime absolutely perfect reed, and put down a whole series of phrasings done the way you would like to be remembered forever. Then the possibility of "cheating" in recordings (which happens all the time anyway) would be endless, you could make sure that what appears on record would always approximate the very best of which you were capable. And -- one would hope! -- would always be in tune.
I imagine a unit like that might sell to a old studio hack or two...
The question for me is: what would happen if you took ALL of these "virtual clarinetists" and integrated them into one "super" virtual performer ... including all the flexibilities of approach that all the individual performers embody. It would be quite interesting to see how this system would handle a given piece of music, especially if you included all the "ethnic" clarinet sounds, Greek, Turkish, Indian, Finnish etc, in the synthesis.
I remember a while back on the internet, can't remember where, seeing a quite serious comment that ran something like -- "All the great saxophonists have a unique sound of their own, but the more you hear the really great clarinetists, the more you notice that they tend to sound the same". At one level this is obviously nonsense, but at another level I can't help wondering whether there is some kind of truth in it -- that there's something about clarinet physics which points to the possibility of "the perfect clarinet sound". I'm just wondering, if integrating increasing numbers of these virtual performers would see a convergence to a particular kind of sound, the Ur-Clarinetist.
Then again, this paragon would presumably sound different in different settings and climates ... there could be an "early" and a "late" version, the young and funky versus the old and reflective ...
One way or another, though, what these guys are doing, is try to make every clarinet composition sound like the same guy under the same conditions, and I'm sure that can't be right.
Last word to Artie Shaw: ""The trouble with composing is that when it's done, it doesn't exist. It's just notes on a piece of paper. Until it's performed. And each performer will stick his own thumbprint on it and change it."
http://www.theatlantic.com/doc/200503/steyn
|
|
Reply To Message
|
|
Author: skygardener
Date: 2008-04-11 05:00
"I mean, if they're actually serious about "compressing" all kinds of recordings in this manner, they could have started with the human voice -- and where would *that* have taken them?"
-- You know, there really is not so much to be scared of with the idea of programs taking the jobs of musician. Part of the great thing about live music is that it is done by real people. Anyone of us has seen weightlifting on TV at one time or another and thought it is amaizing. A person lifting 300kg over their head. But mechines can lift much, much more. Are you impressed when you see a piece of cunstruction equipment lift 500 or 1000kg? I'm not.
|
|
Reply To Message
|
|
Author: monzamess
Date: 2008-04-11 20:56
"You know, there really is not so much to be scared of with the idea of programs taking the jobs of musician. Part of the great thing about live music is that it is done by real people."
Yeah, but, how much of the music you hear is actually live? Think of all the times you hear music, especially the stuff you don't necessarily care to hear (advertisement jingles etc). Much of this stuff is produced as cheaply as possible, which means hiring as few people as possible to produce it, the number of which drops as technology improves.
When computers can write their own code (for the real world, not toy research demonstrations), my career is toast!
|
|
Reply To Message
|
|
Author: skygardener
Date: 2008-04-12 05:18
Then performers should play in an interesting way that is different from 'plastic' music.
|
|
Reply To Message
|
|
The Clarinet Pages
|
 |
.jpg){kind=small}
.jpg){kind=small}
.jpg){kind=small}
.jpg){kind=small}
.jpg){kind=small}
.jpg){kind=small}
