Reminder

classic Classic list List threaded Threaded
23 messages Options
12
Reply | Threaded
Open this post in threaded view
|

Reminder

jpff-2
This is a periodic reminder.  The Csound list welcomes posts from
people with ALL levels of skill, from the newest newbie to the most
serious hacker.  The subject and tenor of the posts vary
dramatically depending on what the current questions are.  Newbies are
sometimes afraid to post because they read discussions about the
incomprehensible deep inner workings, and all they want to know is how
to get a sound to come out of their iBook.  Rest assured that your
question will be answered quickly, and (usually) in a helpful and
courteous manner.  We've all been there.  Please post.

==List Administrator
--
Send bugs reports to this list.
To unsubscribe, send email to [hidden email]
Reply | Threaded
Open this post in threaded view
|

Re: Reminder

Christine L. Myers
Hey everyone!

First, thanks for this "reminder" message that made me feel more comfortable with finally stepping out and asking a few questions.  I'm a newbie to cSound - a major newbie.  I'm a classically-trained pianist, and I've never had the exposure/instruction to electronic music making until recently when I decided to try to learn some stuff on my own.  Partially this comes from buying a year ago my first ever keyboard (before I just only played on acoustic pianos), I bought a Kurzweil PC2X because I got to play a Kurzweil 152i and was blown away with the realistic touch and sound.  The portability and versatility factors definitely sold me into getting caught up to the 21st century of music.  Besides, I never have to have it tuned!  LOL

Anyway, to make a long story short, in my quest to learn how to record on my computer from my keyboard, I ran into cSound and I thought that it would be fun to learn to make music in a totally different way than I've known before.  However, I have never done any computer programming (though I've always wanted to learn - I consider myself a person who learns very quickly, and I've always had a hobby/interest in electronics and mathematics).   I've started with the Tootorial on the csound website, but I still feel pretty clueless.  Should I try to learn about computer programming first, or is that not really necessary?  I did get cSoundAV working and did the first couple of Tootorials, and I got cSoundVST to open but when I put in my orc and sco files, and tell it to Perform it doesn't do anything - it doesn't create a sound file, nothing.  What am I not doing or doing wrong?

So, mainly what I'm asking is if anyone can recommend a good book or website or something that would help me get from the classical-music world to the computer-music world?  I've already done a lot of reading about analogue and digital audio, and the basics of MIDI, but it seems like there are either beginner books that only have about that stuff or advanced books that are way over my head.  I do plan on getting Boulanger's cSound book soon, but I thought I would get something else supplemental for understanding sequencing and synthesis.  I'm planning on buying Cakewalk's Sonar4 HomeStudio soon, and will slowly progress up to Sonar as I have money for the upgrades.  I'm a PC user, with WindowsXP.

I also was recently given by a church a Yamaha DX7, which is in good working condition, so I'm also trying to learn more about synthesis on it. So, if anybody knows a good place to start with it, I'd appreciate it.  I was able to get the manual for it from the internet, but don't know really where to start.

What is funny is, being 30-years old and part of the "computer generation" I consider myself to be pretty knowledgeable and adaptable to computers, the internet, and most forms of technology, but I feel totally like a fish out of water trying to make this leap.  Cursed be all those piano teachers who thought that pop/rock music and keyboards were beneath them and beneath a "true musician"!  I'm kidding, but it is frustrating to feel like I've had a whole world kept from me.  Don't get me wrong, I appreciate my training, and I still love classical music, but I also like pop and rock and folk and jazz and I want to be able to play like that and to make different kinds of music.

So, I'll say thanks ahead of time, and that I really love the list!  I don't understand what you guys are talking about most of the time, but it is awesome that we are all trying to make great music in so many different styles and methods.

Christine Myers



Reply | Threaded
Open this post in threaded view
|

Re: Reminder

Michael Mossey
Hey Christine, welcome to the list!

I don't have much experience helping beginners, so I will let other people
here suggest ideas.

I can say that I approached computer music-making from the opposite
direction: I have a degree in engineering and I'm a professional
programmer, and THEN I got interested in composing. It certainly helps to
know computer programming, because csound code works very much like any
software language. It has some quirks.. is it the best "first language" to
learn? I don't know.. someone else here will probably know.

I can say that if you want to learn programming, Python is a good language
to start, because it is so easy to get programs up and running.. plus
Csound 5 works with Python. However, I can't recommend a book since I only
have Python references lying around, not beginner tutorials.

There is a lot you want to learn.. synthesis is really a topic to itself,
csound is one perspective on that, sequencing is another topic largely to
itself, programming a DX7 is another topic with its own perspective.

Since you can already play piano, a good place to start might be fooling
around on the DX7 with the stock sounds, then starting to modify them.
Then using the sequencer to layer stuff.

Are you the kind of person that wants to dive into a great big pool of
information all at once, or take it a bit at a time?

Mike


<quote who="Christine L. Myers">

> Hey everyone!
>
> First, thanks for this "reminder" message that made me feel more
> comfortable
> with finally stepping out and asking a few questions.  I'm a newbie to
> cSound - a major newbie.  I'm a classically-trained pianist, and I've
> never
> had the exposure/instruction to electronic music making until recently
> when
> I decided to try to learn some stuff on my own.  Partially this comes from
> buying a year ago my first ever keyboard (before I just only played on
> acoustic pianos), I bought a Kurzweil PC2X because I got to play a
> Kurzweil
> 152i and was blown away with the realistic touch and sound.  The
> portability
> and versatility factors definitely sold me into getting caught up to the
> 21st century of music.  Besides, I never have to have it tuned!  LOL
>
> Anyway, to make a long story short, in my quest to learn how to record on
> my
> computer from my keyboard, I ran into cSound and I thought that it would
> be
> fun to learn to make music in a totally different way than I've known
> before.  However, I have never done any computer programming (though I've
> always wanted to learn - I consider myself a person who learns very
> quickly,
> and I've always had a hobby/interest in electronics and mathematics).
> I've
> started with the Tootorial on the csound website, but I still feel pretty
> clueless.  Should I try to learn about computer programming first, or is
> that not really necessary?  I did get cSoundAV working and did the first
> couple of Tootorials, and I got cSoundVST to open but when I put in my orc
> and sco files, and tell it to Perform it doesn't do anything - it doesn't
> create a sound file, nothing.  What am I not doing or doing wrong?
>
> So, mainly what I'm asking is if anyone can recommend a good book or
> website
> or something that would help me get from the classical-music world to the
> computer-music world?  I've already done a lot of reading about analogue
> and
> digital audio, and the basics of MIDI, but it seems like there are either
> beginner books that only have about that stuff or advanced books that are
> way over my head.  I do plan on getting Boulanger's cSound book soon, but
> I
> thought I would get something else supplemental for understanding
> sequencing
> and synthesis.  I'm planning on buying Cakewalk's Sonar4 HomeStudio soon,
> and will slowly progress up to Sonar as I have money for the upgrades.
> I'm
> a PC user, with WindowsXP.
>
> I also was recently given by a church a Yamaha DX7, which is in good
> working
> condition, so I'm also trying to learn more about synthesis on it. So, if
> anybody knows a good place to start with it, I'd appreciate it.  I was
> able
> to get the manual for it from the internet, but don't know really where to
> start.
>
> What is funny is, being 30-years old and part of the "computer generation"
> I
> consider myself to be pretty knowledgeable and adaptable to computers, the
> internet, and most forms of technology, but I feel totally like a fish out
> of water trying to make this leap.  Cursed be all those piano teachers who
> thought that pop/rock music and keyboards were beneath them and beneath a
> "true musician"!  I'm kidding, but it is frustrating to feel like I've had
> a
> whole world kept from me.  Don't get me wrong, I appreciate my training,
> and
> I still love classical music, but I also like pop and rock and folk and
> jazz
> and I want to be able to play like that and to make different kinds of
> music.
>
> So, I'll say thanks ahead of time, and that I really love the list!  I
> don't
> understand what you guys are talking about most of the time, but it is
> awesome that we are all trying to make great music in so many different
> styles and methods.
>
> Christine Myers
>
>
>>
>

--
Send bugs reports to this list.
To unsubscribe, send email to [hidden email]
Reply | Threaded
Open this post in threaded view
|

Shapes

Michael Mossey
In reply to this post by Christine L. Myers
This is a question about computer-assisted composition.

I'm playing with expressive percussion shapes... i.e., beating on a drum
in a regular rhythm, and computing a phrase shape with a beginning and
middle and end. Perhaps starting quiet, getting loud, then quiet again.
Perhaps starting slow tempo, then speeding up, then slowing down.

For example, consider a soft-loud-soft shape. I programmed the amplitudes
as a sinusoidal curve moving from trough to trough. (The curve expressing
the amplitude in db) So how did this sound, subjectively? It wasn't
completely satisfying.. the very rounded top came sounding like the shape
had gotten a bit static during the middle part.

Then I programmed a simple straight line upwards to the middle of the
phrase, and a straight line down. (again straight on a logarithmic scale).
Sounded a little more compelling. Then I modified it to peak such that the
first half length and second half were in the golden ratio. Pretty nice.

Still a little oversimplified sounding since the crescendo and sudden
switch to descrescendo were obvious.

Anyone have ideas for other curves?

Mike

--
Send bugs reports to this list.
To unsubscribe, send email to [hidden email]
Reply | Threaded
Open this post in threaded view
|

Re: Reminder

Steven Yi
In reply to this post by Christine L. Myers
Hi Christine,

Welcome!  It's always nice to hear introductions from new people.

For getting into Csound and Computer Music together, I'd recommend the
book "Virtual Sound" which is structured much like a text book,
illustrating chapter by chapter an aspect of computer music making and
showing how it's done with Csound, slowly building up on the previous
chapters.

To get an initial feel for Csound, I think it's important to
understand the instrument/note paradigm and I'd recommend taking an
orc/sco or csd file with simple instruments in it--preferably a file
you've managed to run with csound and get audio out from--and work
with editing just the note section, editing existing notes, adding and
removing notes, and seeing how that works with the instruments.  Once
you get a feel for that, try the opposite, by keeping the sco the same
and then making very small changes to instruments in the orchestra,
say, changing one opcode line and switching out the opcode, or even
simpler, finding a number in the line and changing the value (but be
careful to keep volumes low on your headphones in case you make a
change that raises the amplitude to ear-piercing levels!).

I don't think you need to know programming prior to working with
Csound, but once you get into it and you start working with making
your own instruments, you will be programming whether you knew it or
not!  But it's not as scary as it might sound, and since it has a
real-world analog to analogue modular synthesizers, it's a little
easier to visualize I think than certain problems that occur in
general programming languages.

Also, I'd highly recommend asking questions here as well as posting
orc/sco or csd files (if not too big) if you need help with anything.
I know I've gotten a great deal of help here in the past and like most
am happy to return the favor.

Good luck and enjoy!
steven


On 3/1/06, Christine L. Myers <[hidden email]> wrote:

> Hey everyone!
>
>  First, thanks for this "reminder" message that made me feel more
> comfortable with finally stepping out and asking a few questions.  I'm a
> newbie to cSound - a major newbie.  I'm a classically-trained pianist, and
> I've never had the exposure/instruction to electronic music making until
> recently when I decided to try to learn some stuff on my own.  Partially
> this comes from buying a year ago my first ever keyboard (before I just only
> played on acoustic pianos), I bought a Kurzweil PC2X because I got to play a
> Kurzweil 152i and was blown away with the realistic touch and sound.  The
> portability and versatility factors definitely sold me into getting caught
> up to the 21st century of music.  Besides, I never have to have it tuned!
> LOL
>
>  Anyway, to make a long story short, in my quest to learn how to record on
> my computer from my keyboard, I ran into cSound and I thought that it would
> be fun to learn to make music in a totally different way than I've known
> before.  However, I have never done any computer programming (though I've
> always wanted to learn - I consider myself a person who learns very quickly,
> and I've always had a hobby/interest in electronics and mathematics).   I've
> started with the Tootorial on the csound website, but I still feel pretty
> clueless.  Should I try to learn about computer programming first, or is
> that not really necessary?  I did get cSoundAV working and did the first
> couple of Tootorials, and I got cSoundVST to open but when I put in my orc
> and sco files, and tell it to Perform it doesn't do anything - it doesn't
> create a sound file, nothing.  What am I not doing or doing wrong?
>
>  So, mainly what I'm asking is if anyone can recommend a good book or
> website or something that would help me get from the classical-music world
> to the computer-music world?  I've already done a lot of reading about
> analogue and digital audio, and the basics of MIDI, but it seems like there
> are either beginner books that only have about that stuff or advanced books
> that are way over my head.  I do plan on getting Boulanger's cSound book
> soon, but I thought I would get something else supplemental for
> understanding sequencing and synthesis.  I'm planning on buying Cakewalk's
> Sonar4 HomeStudio soon, and will slowly progress up to Sonar as I have money
> for the upgrades.  I'm a PC user, with WindowsXP.
>
>  I also was recently given by a church a Yamaha DX7, which is in good
> working condition, so I'm also trying to learn more about synthesis on it.
> So, if anybody knows a good place to start with it, I'd appreciate it.  I
> was able to get the manual for it from the internet, but don't know really
> where to start.
>
>  What is funny is, being 30-years old and part of the "computer generation"
> I consider myself to be pretty knowledgeable and adaptable to computers, the
> internet, and most forms of technology, but I feel totally like a fish out
> of water trying to make this leap.  Cursed be all those piano teachers who
> thought that pop/rock music and keyboards were beneath them and beneath a
> "true musician"!  I'm kidding, but it is frustrating to feel like I've had a
> whole world kept from me.  Don't get me wrong, I appreciate my training, and
> I still love classical music, but I also like pop and rock and folk and jazz
> and I want to be able to play like that and to make different kinds of
> music.
>
>  So, I'll say thanks ahead of time, and that I really love the list!  I
> don't understand what you guys are talking about most of the time, but it is
> awesome that we are all trying to make great music in so many different
> styles and methods.
>
>  Christine Myers
>
>
> >
> >
>
>
--
Send bugs reports to this list.
To unsubscribe, send email to [hidden email]
Reply | Threaded
Open this post in threaded view
|

Re: Reminder

jpff
I was about to say much the same as Steven.  My experience may be of
little resonance as my background is mathematics, and programming
since a long time ago.  I found csound terrifying wen I first found
it, and could not work out what to do with it.  There was no such
mailing list then and I struggled.  What finally got be going was
trying the examples which were distributed with csound until I found
a sound I really liked (actually Risset's marimba) and then I played
with it, initially fairly randomly.
  The books are easier to follow when you have an aim, in my opinion.

However, welcome to the list.  Keep asking until you get answers!
==John ffitch
--
Send bugs reports to this list.
To unsubscribe, send email to [hidden email]
Reply | Threaded
Open this post in threaded view
|

Re: Reminder

luis jure
In reply to this post by Christine L. Myers
el Wed, 1 Mar 2006 21:12:39 -0500
"Christine L. Myers" <[hidden email]> escribi├│:


> Should I try to learn about computer
> programming first, or is that not really necessary?  

in my opinion you can learn csound, and be pretty good at it too, without
having to learn computer programming languages first. although it can be a
bonus, of course, specially if you're interested in algorithmic composition.
it seems python is a good choice. but in your place, i'd concentrate on
csound first.

and the only important piece of advice i have to offer: edit your csound
files with a simple text editor, and compile from the command line. when you
understand what's happening, you can choose some front-end or GUI if you
think you'll be more productive that way.


> So, mainly what I'm asking is if anyone can recommend a good book or
> website or something that would help me get from the classical-music world
> to the computer-music world?

i strongly recommend this book:

Dodge, Charles & Thomas Jerse. 1997. Computer Music. Synthesis, composition,
and performance. Schirmer Books, 2ª ed.

it covers basically everything, it's very clear, and authored by one of the
best composers of computer music.

good luck,

lj

--
Send bugs reports to this list.
To unsubscribe, send email to [hidden email]
Reply | Threaded
Open this post in threaded view
|

OT: Music and Artificial Neural Networks

Dale Stewart
Has anyone here used Artificial Neural Network technology to compose or
analyze music in some way?  I am currently beginning to read the book Music
and Connectionism (edited by Peter M. Todd and D. Gareth Loy, 1991, ISBN
0-262-20081-3).  

At one point, around the time period this book was written, I was exposed to
some of the work of Teuvo Kohonen, and I was fascinated by it.  This book is
a collection of papers, of which he is one of the authors.  "A Nonheuristic
Automatic Composing Method" is the title of his paper.  Can anyone relate
any experiences with this?  What is the current state of the art in this
area?  Thanks in advance for any replies.



--
Send bugs reports to this list.
To unsubscribe, send email to [hidden email]
Reply | Threaded
Open this post in threaded view
|

Re: OT: Music and Artificial Neural Networks

Jeff Harrington
I used a Kohonen net I built to generate melodies experimentally a while
back.  At the time,
though, I wasn't convinced as to what I had actually trained the net to
produce.  I trained it using
4 or 5 examples of short melodies and it produced a short melody within
the same style.

I gave it dyads - pitch/duration in a sequence.  And it spat out a
sequence of dyads.

I wrote the whole thing in Icon, a great parsing language.

Jeff
http://jeffharrington.org


Dale Stewart wrote:

>Has anyone here used Artificial Neural Network technology to compose or
>analyze music in some way?  I am currently beginning to read the book Music
>and Connectionism (edited by Peter M. Todd and D. Gareth Loy, 1991, ISBN
>0-262-20081-3).  
>
>At one point, around the time period this book was written, I was exposed to
>some of the work of Teuvo Kohonen, and I was fascinated by it.  This book is
>a collection of papers, of which he is one of the authors.  "A Nonheuristic
>Automatic Composing Method" is the title of his paper.  Can anyone relate
>any experiences with this?  What is the current state of the art in this
>area?  Thanks in advance for any replies.
>
>
>
>  
>

--
Send bugs reports to this list.
To unsubscribe, send email to [hidden email]
Reply | Threaded
Open this post in threaded view
|

Re: OT: Music and Artificial Neural Networks

Dale Stewart
Jeff,

Thanks for the tip.  Found the Icon language here:

http://www.cs.arizona.edu/icon/index.htm

Is there any chance of posting your implementation, or is it already
available somewhere?

Thanks,
Dale

-----Original Message-----
From: Jeff Harrington [mailto:[hidden email]]
Sent: Thursday, March 02, 2006 8:37 AM
To: [hidden email]
Subject: Re: [Csnd] OT: Music and Artificial Neural Networks

I used a Kohonen net I built to generate melodies experimentally a while
back.  At the time,
though, I wasn't convinced as to what I had actually trained the net to
produce.  I trained it using
4 or 5 examples of short melodies and it produced a short melody within
the same style.

I gave it dyads - pitch/duration in a sequence.  And it spat out a
sequence of dyads.

I wrote the whole thing in Icon, a great parsing language.

Jeff
http://jeffharrington.org


Dale Stewart wrote:

>Has anyone here used Artificial Neural Network technology to compose or
>analyze music in some way?  I am currently beginning to read the book Music
>and Connectionism (edited by Peter M. Todd and D. Gareth Loy, 1991, ISBN
>0-262-20081-3).  
>
>At one point, around the time period this book was written, I was exposed
to
>some of the work of Teuvo Kohonen, and I was fascinated by it.  This book
is
>a collection of papers, of which he is one of the authors.  "A Nonheuristic
>Automatic Composing Method" is the title of his paper.  Can anyone relate
>any experiences with this?  What is the current state of the art in this
>area?  Thanks in advance for any replies.
>
>
>
>  
>

--
Send bugs reports to this list.
To unsubscribe, send email to [hidden email]

--
Send bugs reports to this list.
To unsubscribe, send email to [hidden email]
Reply | Threaded
Open this post in threaded view
|

Re: OT: Music and Artificial Neural Networks

Anthony Kozar
In reply to this post by Jeff Harrington
I attempted something similar several years ago but using Hebbian networks
instead (unsupervised learning).  I feed Gregorian chant melodies into a
network by giving it a sequence of tuples where each tuple included the
current pitch and N previous pitches.  I hoped that the network would learn
some of the characteristic melodic patterns of the chant and then be able to
reproduce authentic sounding sequences for new melodies.

It really did not work very well and a simple markov chain implementation
worked better (which is what I was imitating with the network anyways).

My code was written in Common Lisp, making it easy to experiment in an
interactive manner.  I would be happy to share it with anyone who would find
it useful.

Anthony Kozar
anthonykozar AT sbcglobal DOT net

Jeff Harrington wrote on 3/2/06 9:36 AM:

> I used a Kohonen net I built to generate melodies experimentally a while
> back.  At the time, though, I wasn't convinced as to what I had actually
> trained the net to produce.  I trained it using 4 or 5 examples of short
> melodies and it produced a short melody within the same style.

> Dale Stewart wrote:
>
>> Has anyone here used Artificial Neural Network technology to compose or
>> analyze music in some way?  

--
Send bugs reports to this list.
To unsubscribe, send email to [hidden email]
Reply | Threaded
Open this post in threaded view
|

Re: OT: Music and Artificial Neural Networks

Dale Stewart
Anthony,

Thanks for your response.  I would be very interested in seeing what you
did, in whatever way you would like to make it available.

Thanks,
Dale

-----Original Message-----
From: Anthony Kozar [mailto:[hidden email]]
Sent: Thursday, March 02, 2006 10:23 AM
To: [hidden email]
Subject: Re: [Csnd] OT: Music and Artificial Neural Networks

I attempted something similar several years ago but using Hebbian networks
instead (unsupervised learning).  I feed Gregorian chant melodies into a
network by giving it a sequence of tuples where each tuple included the
current pitch and N previous pitches.  I hoped that the network would learn
some of the characteristic melodic patterns of the chant and then be able to
reproduce authentic sounding sequences for new melodies.

It really did not work very well and a simple markov chain implementation
worked better (which is what I was imitating with the network anyways).

My code was written in Common Lisp, making it easy to experiment in an
interactive manner.  I would be happy to share it with anyone who would find
it useful.

Anthony Kozar
anthonykozar AT sbcglobal DOT net

Jeff Harrington wrote on 3/2/06 9:36 AM:

> I used a Kohonen net I built to generate melodies experimentally a while
> back.  At the time, though, I wasn't convinced as to what I had actually
> trained the net to produce.  I trained it using 4 or 5 examples of short
> melodies and it produced a short melody within the same style.

> Dale Stewart wrote:
>
>> Has anyone here used Artificial Neural Network technology to compose or
>> analyze music in some way?  

--
Send bugs reports to this list.
To unsubscribe, send email to [hidden email]

--
Send bugs reports to this list.
To unsubscribe, send email to [hidden email]
Reply | Threaded
Open this post in threaded view
|

Re: OT: Music and Artificial Neural Networks

Jeff Harrington
In reply to this post by Dale Stewart
Dale, it was written for my Amiga 3000 and I would have to turn it on to get
the source...  :-P  Actually I think I was feeding data through into an
old C++ Kohonen
net example I got online.   Doubt it would help even if I could find it.

Not saying Icon is a good language for anything these days, but it is
interesting
as almost a kind of Ruby precursor.

As Anthony suggests, I'm not sure that time series data is a good fit for a
neural net solution.  Markov processes definitely sound more appropriate and
I've used them much more (even in my 1st symphony).    My appreciation of
my experiments is that I was basically telling the net to generate a
large number
'like' the large number I had trained it to produce.

Jeff
http://jeffharrington.org

Dale Stewart wrote:

>Jeff,
>
>Thanks for the tip.  Found the Icon language here:
>
>http://www.cs.arizona.edu/icon/index.htm
>
>Is there any chance of posting your implementation, or is it already
>available somewhere?
>
>Thanks,
>Dale
>
>-----Original Message-----
>From: Jeff Harrington [mailto:[hidden email]]
>Sent: Thursday, March 02, 2006 8:37 AM
>To: [hidden email]
>Subject: Re: [Csnd] OT: Music and Artificial Neural Networks
>
>I used a Kohonen net I built to generate melodies experimentally a while
>back.  At the time,
>though, I wasn't convinced as to what I had actually trained the net to
>produce.  I trained it using
>4 or 5 examples of short melodies and it produced a short melody within
>the same style.
>
>I gave it dyads - pitch/duration in a sequence.  And it spat out a
>sequence of dyads.
>
>I wrote the whole thing in Icon, a great parsing language.
>
>Jeff
>http://jeffharrington.org
>
>
>Dale Stewart wrote:
>
>  
>
>>Has anyone here used Artificial Neural Network technology to compose or
>>analyze music in some way?  I am currently beginning to read the book Music
>>and Connectionism (edited by Peter M. Todd and D. Gareth Loy, 1991, ISBN
>>0-262-20081-3).  
>>
>>At one point, around the time period this book was written, I was exposed
>>    
>>
>to
>  
>
>>some of the work of Teuvo Kohonen, and I was fascinated by it.  This book
>>    
>>
>is
>  
>
>>a collection of papers, of which he is one of the authors.  "A Nonheuristic
>>Automatic Composing Method" is the title of his paper.  Can anyone relate
>>any experiences with this?  What is the current state of the art in this
>>area?  Thanks in advance for any replies.
>>
>>
>>
>>
>>
>>    
>>
>
>  
>

--
Send bugs reports to this list.
To unsubscribe, send email to [hidden email]
Reply | Threaded
Open this post in threaded view
|

Re: OT: Music and Artificial Neural Networks

Dale Stewart
Jeff,

Ok, that has been a while, hasn't it! :-)

I will forge ahead into the book and see what I can come up with.  My most
fluent programming language is C++ at the moment, but I have noticed a lot
of folks on here using Python, Ruby, Lisp, etc...  I would like to take some
time to learn more about those at some point.

If I come up with anything interesting, I will let you know.  I will also
check into Markov processes in composition as well.  

The appeal of neural networks to me is that it attempts to model brain
activity in a more direct way than any other method.  It seems the biggest
limitation of this is that the number of neurons that you can successfully
model at a time is fairly small in comparison to the real thing.  And, as
with any automatic composition method, you have to give it some input to
guide it into producing something "good", and then make a judgment of
whether it is "good" when you're done.  So, it is still a very human
activity in the long run.  So I see it as a tool to help generate ideas more
rapidly, rather than a total solution to the composition problem.

I am still very new to music theory as well, so any suggestions for
fundamental texts that would help me along this path would be greatly
appreciated.  My family and I have been into violin, viola, and cello for
that past 2-3 years now, and my interest in music is growing rapidly.  My
background is in engineering and software development, so I am coming at it
from that angle.  I am personally learning to play cello, trying to learn to
play the Bach Suite I Prelude (BWV 1007) at the moment.  I want to learn as
much as I can about composition as well, as I see it as a similar activity
to programming, in some sense, and I enjoy being creative in that way.

Thanks,
Dale

-----Original Message-----
From: Jeff Harrington [mailto:[hidden email]]
Sent: Thursday, March 02, 2006 10:35 AM
To: [hidden email]
Subject: Re: [Csnd] OT: Music and Artificial Neural Networks

Dale, it was written for my Amiga 3000 and I would have to turn it on to get
the source...  :-P  Actually I think I was feeding data through into an
old C++ Kohonen
net example I got online.   Doubt it would help even if I could find it.

Not saying Icon is a good language for anything these days, but it is
interesting
as almost a kind of Ruby precursor.

As Anthony suggests, I'm not sure that time series data is a good fit for a
neural net solution.  Markov processes definitely sound more appropriate and
I've used them much more (even in my 1st symphony).    My appreciation of
my experiments is that I was basically telling the net to generate a
large number
'like' the large number I had trained it to produce.

Jeff
http://jeffharrington.org

Dale Stewart wrote:

>Jeff,
>
>Thanks for the tip.  Found the Icon language here:
>
>http://www.cs.arizona.edu/icon/index.htm
>
>Is there any chance of posting your implementation, or is it already
>available somewhere?
>
>Thanks,
>Dale
>
>-----Original Message-----
>From: Jeff Harrington [mailto:[hidden email]]
>Sent: Thursday, March 02, 2006 8:37 AM
>To: [hidden email]
>Subject: Re: [Csnd] OT: Music and Artificial Neural Networks
>
>I used a Kohonen net I built to generate melodies experimentally a while
>back.  At the time,
>though, I wasn't convinced as to what I had actually trained the net to
>produce.  I trained it using
>4 or 5 examples of short melodies and it produced a short melody within
>the same style.
>
>I gave it dyads - pitch/duration in a sequence.  And it spat out a
>sequence of dyads.
>
>I wrote the whole thing in Icon, a great parsing language.
>
>Jeff
>http://jeffharrington.org
>
>
>Dale Stewart wrote:
>
>  
>
>>Has anyone here used Artificial Neural Network technology to compose or
>>analyze music in some way?  I am currently beginning to read the book
Music

>>and Connectionism (edited by Peter M. Todd and D. Gareth Loy, 1991, ISBN
>>0-262-20081-3).  
>>
>>At one point, around the time period this book was written, I was exposed
>>    
>>
>to
>  
>
>>some of the work of Teuvo Kohonen, and I was fascinated by it.  This book
>>    
>>
>is
>  
>
>>a collection of papers, of which he is one of the authors.  "A
Nonheuristic

>>Automatic Composing Method" is the title of his paper.  Can anyone relate
>>any experiences with this?  What is the current state of the art in this
>>area?  Thanks in advance for any replies.
>>
>>
>>
>>
>>
>>    
>>
>
>  
>

--
Send bugs reports to this list.
To unsubscribe, send email to [hidden email]

--
Send bugs reports to this list.
To unsubscribe, send email to [hidden email]
Reply | Threaded
Open this post in threaded view
|

Re: OT: Music and Artificial Neural Networks

Anthony Kozar
In reply to this post by Dale Stewart
I've uploaded the source code and some midi examples along with a PDF of the
project report that I wrote (This was a school project).  I dubbed it "NN
Chanter" and the files are at:

http://briefcase.yahoo.com/bc/akozar@.../lst?.dir=/NNChanter

All are welcome to download and use however you wish.  Consider the code
public domain -- the hebbian-net and matrix math units are hopefully fairly
reusable.  Everything is written in generic Common Lisp except for one file
-- chant-player.lisp -- that uses Common Music to convert my note lists to
Midi.

Let me know if you find any use for it or have any questions.

Anthony Kozar
anthonykozar AT sbcglobal DOT net


Dale Stewart wrote on 3/2/06 11:31 AM:

> Thanks for your response.  I would be very interested in seeing what you
> did, in whatever way you would like to make it available.

--
Send bugs reports to this list.
To unsubscribe, send email to [hidden email]
Reply | Threaded
Open this post in threaded view
|

Re: OT: Music and Artificial Neural Networks

Dale Stewart
Thanks for that.  I actually had to enter the address without the
"/lst?.dir=/NNChanter" part to find it.  So far I've downloaded the files,
scanned over the paper, and listened to your inputs and outputs.  There is
still some musical viability in the output, I think, although it doesn't
seem to match the style of the inputs extremely well.  I will study the
implementation when I get a chance and see where I can take it.  I think you
have done some outstanding work here, though.

-----Original Message-----
From: Anthony Kozar [mailto:[hidden email]]
Sent: Thursday, March 02, 2006 1:00 PM
To: [hidden email]
Subject: Re: [Csnd] OT: Music and Artificial Neural Networks

I've uploaded the source code and some midi examples along with a PDF of the
project report that I wrote (This was a school project).  I dubbed it "NN
Chanter" and the files are at:

http://briefcase.yahoo.com/bc/akozar@.../lst?.dir=/NNChanter

All are welcome to download and use however you wish.  Consider the code
public domain -- the hebbian-net and matrix math units are hopefully fairly
reusable.  Everything is written in generic Common Lisp except for one file
-- chant-player.lisp -- that uses Common Music to convert my note lists to
Midi.

Let me know if you find any use for it or have any questions.

Anthony Kozar
anthonykozar AT sbcglobal DOT net


Dale Stewart wrote on 3/2/06 11:31 AM:

> Thanks for your response.  I would be very interested in seeing what you
> did, in whatever way you would like to make it available.

--
Send bugs reports to this list.
To unsubscribe, send email to [hidden email]

--
Send bugs reports to this list.
To unsubscribe, send email to [hidden email]
Reply | Threaded
Open this post in threaded view
|

Re: OT: Music and Artificial Neural Networks

Anthony Kozar
Thanks very much for the feedback.  Any musicality of the output might just
be a result of "hard-wiring" in the Dorian scale that the chants used.
There is no possibility of the network generating a note outside of that
scale.

But I am glad that you find it interesting and maybe you will be able to
improve on the idea!

Thanks.

Anthony Kozar
anthonykozar AT sbcglobal DOT net


Dale Stewart wrote on 3/2/06 2:38 PM:

> Thanks for that.  I actually had to enter the address without the
> "/lst?.dir=/NNChanter" part to find it.  So far I've downloaded the files,
> scanned over the paper, and listened to your inputs and outputs.  There is
> still some musical viability in the output, I think, although it doesn't
> seem to match the style of the inputs extremely well.  I will study the
> implementation when I get a chance and see where I can take it.  I think you
> have done some outstanding work here, though.

--
Send bugs reports to this list.
To unsubscribe, send email to [hidden email]
Reply | Threaded
Open this post in threaded view
|

Re: Shapes

Aidan Collins
In reply to this post by Michael Mossey
I always get frustrated when people suggest using different software
as a response to a question, but I can't help thinking you should try
Cecilia. It lets you draw such curves graphically (breakpoints and
whatnot), so you could try out a number of them in no time at all.

otherwise, what about trying the sinusoid again (or maybe a triangle
wave) but have the frequency vary a little bit, maybe slowing down
around the peak amplitude to stretch out the climax?

On 3/1/06, Michael Mossey <[hidden email]> wrote:

> This is a question about computer-assisted composition.
>
> I'm playing with expressive percussion shapes... i.e., beating on a drum
> in a regular rhythm, and computing a phrase shape with a beginning and
> middle and end. Perhaps starting quiet, getting loud, then quiet again.
> Perhaps starting slow tempo, then speeding up, then slowing down.
>
> For example, consider a soft-loud-soft shape. I programmed the amplitudes
> as a sinusoidal curve moving from trough to trough. (The curve expressing
> the amplitude in db) So how did this sound, subjectively? It wasn't
> completely satisfying.. the very rounded top came sounding like the shape
> had gotten a bit static during the middle part.
>
> Then I programmed a simple straight line upwards to the middle of the
> phrase, and a straight line down. (again straight on a logarithmic scale).
> Sounded a little more compelling. Then I modified it to peak such that the
> first half length and second half were in the golden ratio. Pretty nice.
>
> Still a little oversimplified sounding since the crescendo and sudden
> switch to descrescendo were obvious.
>
> Anyone have ideas for other curves?
>
> Mike
>
> --
> Send bugs reports to this list.
> To unsubscribe, send email to [hidden email]
>
--
Send bugs reports to this list.
To unsubscribe, send email to [hidden email]
Reply | Threaded
Open this post in threaded view
|

Re: Shapes

Steven Yi
In reply to this post by Michael Mossey
Hi Michael,

This doesn't quite address the question directly, but I just wanted to
note something from my experience with triangular envelopes.  I use
linear triangular envelopes as a basis for most of what I do, mostly
on the single note level and not at the phrase.  I've found that the
duration of notes or the scale of ideas deeply affects the experience
of the triangle; if the duration is shorter, the slope of the triangle
is steeper, and if the duration is longer, the slope is more gradual,
given that the start and peak values are the same.

Perhaps obvious, but it was something that affected how I worked with
triangular enveloped sounds and has become a very important quality to
me.  I did experiment with half-sine's and various other shapes as
well but eventually settled on the linear triangular envelope.  I
think keeping in mind the slope and that if you're working with
something that is duration dependent might help to guide finding the
shape that will satisfy your ear.

Hope that helps!
steven



On 3/1/06, Michael Mossey <[hidden email]> wrote:

> This is a question about computer-assisted composition.
>
> I'm playing with expressive percussion shapes... i.e., beating on a drum
> in a regular rhythm, and computing a phrase shape with a beginning and
> middle and end. Perhaps starting quiet, getting loud, then quiet again.
> Perhaps starting slow tempo, then speeding up, then slowing down.
>
> For example, consider a soft-loud-soft shape. I programmed the amplitudes
> as a sinusoidal curve moving from trough to trough. (The curve expressing
> the amplitude in db) So how did this sound, subjectively? It wasn't
> completely satisfying.. the very rounded top came sounding like the shape
> had gotten a bit static during the middle part.
>
> Then I programmed a simple straight line upwards to the middle of the
> phrase, and a straight line down. (again straight on a logarithmic scale).
> Sounded a little more compelling. Then I modified it to peak such that the
> first half length and second half were in the golden ratio. Pretty nice.
>
> Still a little oversimplified sounding since the crescendo and sudden
> switch to descrescendo were obvious.
>
> Anyone have ideas for other curves?
>
> Mike
>
> --
> Send bugs reports to this list.
> To unsubscribe, send email to [hidden email]
>
--
Send bugs reports to this list.
To unsubscribe, send email to [hidden email]
Reply | Threaded
Open this post in threaded view
|

Re: Shapes

Michael Mossey
Hi Steven,

Yes, I think it is relevant, because composition can be self-similar..
perhaps I could use the same envelope on notes as I use on phrases.

Do you always have the triangle peak in the middle time-wise, or have you
ever tried having it peak 2/3 of the way through, or at the Golden ratio,
or something?

Mike



<quote who="Steven Yi">

> Hi Michael,
>
> This doesn't quite address the question directly, but I just wanted to
> note something from my experience with triangular envelopes.  I use
> linear triangular envelopes as a basis for most of what I do, mostly
> on the single note level and not at the phrase.  I've found that the
> duration of notes or the scale of ideas deeply affects the experience
> of the triangle; if the duration is shorter, the slope of the triangle
> is steeper, and if the duration is longer, the slope is more gradual,
> given that the start and peak values are the same.
>
> Perhaps obvious, but it was something that affected how I worked with
> triangular enveloped sounds and has become a very important quality to
> me.  I did experiment with half-sine's and various other shapes as
> well but eventually settled on the linear triangular envelope.  I
> think keeping in mind the slope and that if you're working with
> something that is duration dependent might help to guide finding the
> shape that will satisfy your ear.
>
> Hope that helps!
> steven
>
>
>
> On 3/1/06, Michael Mossey <[hidden email]> wrote:
>> This is a question about computer-assisted composition.
>>
>> I'm playing with expressive percussion shapes... i.e., beating on a drum
>> in a regular rhythm, and computing a phrase shape with a beginning and
>> middle and end. Perhaps starting quiet, getting loud, then quiet again.
>> Perhaps starting slow tempo, then speeding up, then slowing down.
>>
>> For example, consider a soft-loud-soft shape. I programmed the
>> amplitudes
>> as a sinusoidal curve moving from trough to trough. (The curve
>> expressing
>> the amplitude in db) So how did this sound, subjectively? It wasn't
>> completely satisfying.. the very rounded top came sounding like the
>> shape
>> had gotten a bit static during the middle part.
>>
>> Then I programmed a simple straight line upwards to the middle of the
>> phrase, and a straight line down. (again straight on a logarithmic
>> scale).
>> Sounded a little more compelling. Then I modified it to peak such that
>> the
>> first half length and second half were in the golden ratio. Pretty nice.
>>
>> Still a little oversimplified sounding since the crescendo and sudden
>> switch to descrescendo were obvious.
>>
>> Anyone have ideas for other curves?
>>
>> Mike
>>
>> --
>> Send bugs reports to this list.
>> To unsubscribe, send email to [hidden email]
>>
> --
> Send bugs reports to this list.
> To unsubscribe, send email to [hidden email]
>

--
Send bugs reports to this list.
To unsubscribe, send email to [hidden email]
12