EMUSIC-L Digest Volume 50, Issue 16 This issue's topics: free Xwin patch editor+librarian It's all about Voice Paths (TM) Machine Models of Music - Any comments? (6 messages) Your EMUSIC-L Digest moderator is Joe McMahon . You may subscribe to EMUSIC-L by sending mail to listserv@american.edu with the line "SUB EMUSIC-L your name" as the text. The EMUSIC-L archive is a service of SunSite (sunsite.unc.edu) at the University of North Carolina. ------------------------------------------------------------------------ Date: Wed, 31 Mar 1993 17:13:50 GMT From: Greg Lee Subject: free Xwin patch editor+librarian Organization: University of Hawaii The source code for an X Windows version of glib is available by ftp from tsx-11.mit.edu as /pub/linux/sources/usr.bin/glib19.tar.z (compressed with gzip). To compile it requires Xview+Slingshot. Unless you happen to be running Linux and Savolainen's sound driver it will require some coding of the i/o routines to get sound out of it or to upload to a synth. It has support now for only the SoundBlaster and the Kawaii K1, though adding back in support for other synths that glib previously handled should not be too hard. (The earlier curses version of glib is on ucsd.edu in /midi/software/misc/glib2.tar.Z.) -- Greg Lee ------------------------------ Date: Sat, 13 Mar 1993 10:39:32 +0000 From: Nick Rothwell Subject: It's all about Voice Paths (TM) Something's been nagging at the back of my mind ever since I had to knock up all that virtual master keyboard code in MAX for that modern dance thing last year. I posted something a few months ago about how most keyboard instruments these days are stop-and-go: master keyboards that let you reconfigure an entire arsenal of synth modules, and yet which kill all voices everywhere if you attempt to just patch change the poor Matrix-1000 in the corner. Synthesisers which kill all voices when you send a patch change on one of the single tone channels, or which only let you flip effects by changing an entire multimode (thus losing the settings, patches, controller values of every single reception channel). So far, so bad. I'm probably about to start hacking up a MAX control system for the Fostex MIDI mixers. Yes, the editing tablet is fine, but the mixers are designed to take program changes for total reconfiguration. This isn't good enough: what I want is some way of having each mixer hold about ten independent patches at once: one for each channel, and perhaps a couple for send/return stuff. Multitimbral control, in effect, so that channels can be reconfigured independently. I sort-of had most of this designed in my head several weeks ago (the idea being that there would be master patch changes which would call up subgroup patch changes, perhaps with programmable crossover fade times), but somehow I wasn't totally pleased with the idea. It seemed too complicated. Then I started hacking into the Wavestation effects programs with some MAX and C externals, trying to set up an effects journaliser and librarian so that I could change effects programs on the fly by replaying edit sequences. Somehow a step forward, but adding too much control complexity: an abstraction with too many control points, like the MIDI mixer thing. Now, finally, I think I've got my head around exactly what abstraction I'm trying to build. Sort-of came to me at the same time as a set of aching hamstrings, the result of a weekend intensive course on T'ai Chi, Chi Gung and anatomical principles of movement, throwing people across the room with one finger, that kind of stuff. Lots of theory about kinetic chains of movement, which turned on a light bulb somewhere: I suddenly saw a connection between the stuff we were doing with acupressure points and the sort of software I was trying to build to compose with, and control, the MIDI rig. The problem is one of connectivity, between performance input (keys played, whatever) and the stuff which comes out of the speakers. It's not right to think in terms of program changes and presets except for the simplest monotimbral synths and effects units, and multitimbral setups and multichannel effects processors are a distraction to the abstraction I'm trying to describe. Configuring a keyboard is not a process of calling up patches: wrong model. It's a process of establishing a path between a subset of the control surfaces (keys, wheels etc.) and a collection of final sounds (by "final" I mean at the stereo output: effects and all). At any one time, the configuration of a keyboard rig is not a selection of multipatches and split keyboards: it's a collection of active voice paths, mutually independent lines of control from bits of keyboard to distinct elements of the music. "Voice path" is a vague term, deliberately: it's not just MIDI, or just audio: it's both. One end of the voice path is keys and wheels generating MIDI. That goes through transposition, splitting, maybe the odd MIDI processor like an arpeggiator. Somewhere it goes across the boundary from control signal to audio (in the the synthesiser); and somewhere it goes across the boundary from multitimbral audio (a pure sound, sans effects) to compromised audio (sharing an effects processor). Oh, and there's compromise sometimes in the performance architecture, in that (for example) the elements of a multiset cannot be selected independently (e.g. D-550). So, the picture to get in your mind is that a keyboard rig is something which lets you build simultaneous, independent voice paths. What then becomes important for live performance is how flexibly you can create and remove voice paths without them interfering with each other. If the set of voice paths is constant for each "song", then a Roland master keyboard and some conventional synths will keep you happy, since the abstractions of voice path and keyboard setup, multiset, whatever happen to coincide. If you want to change voice paths dynamically, then you now see why Roland master keyboards and the multiset mentality are so crippled. Try to alter one path, and immediately every other path is destroyed. In an ideal world, all voice paths would be independent. Without truly multitimbral effects processors, and MUCH better designed MIDI control software in our synthesisers, we can't do that yet; I consider this to be the reason for most of my frustration when composing and performing with MIDI: I spend so much time making sure that voice paths are not violated by each other. (MIDI sequencers are assemblers, note: they work in the machine code of their application area, viz. MIDI events. What we need is voice path compilers.) And now I know what I want to do in MAX. A patch change isn't something which selects patches. It's something which establishes a voice path, altering several devices at once, or rather, *parts* of several devices (patch change on a synth, some SysEx setup of the onboard effects, some controller messages to change one channel on one of the MIDI mixers). Hmm. Yeah. I reckon 10 or 15 megabytes of MAX code should do it... Nick Rothwell | cassiel@cassiel.demon.co.uk CASSIEL Contemporary Music/Dance | cassiel@cix.compulink.co.uk ------------------------------ Date: Fri, 5 Mar 1993 00:00:51 -0600 From: Bruce Colton Subject: Re: Machine Models of Music - Any comments? There was some significant work done by Kahonen using Neural Networks. I attended a reception at the Joint Conference on Neural Networks at which the entire evening was accompanied by Kahonen's Neural generated music from a Bach trained network. Found it to be very interesting and since you indicate your knowledge of Minsky you are probably also aware of Kahonen who I believe is Finish. At any rate i just thought that I would mention it in case you were interested. ------------------------------ Date: Fri, 5 Mar 1993 00:28:00 PST From: BISHOP_M@PLU.BITNET Subject: Re: Machine Models of Music - Any comments? I am very interested in hearing some of this music. Where could I find it? I have a friend who is working on AI and would be interested as well. Does it sound intersting? Is it profound in any way? Is there any way to get it to interact in real-time with human input? Does it use MIDI as it's interface? Mike Bishop BISHOP_M@PLU.BITNET ------------------------------ Date: Fri, 5 Mar 1993 08:10:18 PST From: Leigh Smith Subject: Machine Models of Music - Any comments? Hi All, Has anyone had a chance to look at 'Machine Models of Music: Minsky to Mozart' from the MIT Press yet? If so, can they give a bit of a review or minimally, a list of the papers presented? I haven't seen it appear in Piet's reference list at RUU yet. Thanks -- =--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--= | Leigh Smith _--_|\ Email: leigh@psychok.DIALix.oz.au | | / \ Tel (H): +61-9-450-2014 | | Perth --> *_.--._/ Perth, Western Australia | | v | =--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--= ------------------------------ Date: Fri, 5 Mar 1993 17:20:09 GMT From: Andy Farnell Subject: Re: Machine Models of Music - Any comments? > Has anyone had a chance to look at 'Machine Models of Music: Minsky to > Mozart' from the MIT Press yet? If so, can they give a bit of a review or > minimally, a list of the papers presented? I haven't seen it appear in Piet's > reference list at RUU yet. > This sounds VERY interesting to me Leigh, Does anyone have an ISBN ref. Might this work include AI applications to orchestration and texture generation? Andy Farnell Computing and Cognition Bournemouth University England ------------------------------ Date: Fri, 5 Mar 1993 18:21:18 GMT From: JOEL STERN Subject: RE: MACHINE MODELS OF MUSIC - > Has anyone had a chance to look at 'Machine Models of Music: Minsky to > Mozart' from the MIT Press yet? If so, can they give a bit of a review or > minimally, a list of the papers presented? I haven't seen it appear in Piet's > reference list at RUU yet. > This sounds VERY interesting to me Leigh, Does anyone have an ISBN ref. Might this work include AI applications to orchestration and texture generation? Andy Farnell Computing and Cognition Bournemouth University England ------------------------------------------ THIS IS A REPLY TO THE ABOVE MESSAGE SUBJECT OF THE REPLY: REPLY ------------------------------------------ Andy Farnell writes re "Machine Models of Music": >This sounds VERY interesting to me Leigh, Does anyone have an IS >Might this work include AI applications to orchestration and tex >generation? ISBN is 0-262-19319-1 Joel Stern stern@mail.loc.gov ------------------------------ Date: Sat, 6 Mar 1993 16:02:46 PST From: Leigh Smith Subject: Machine Models info My question I posed before was: Has anyone had a chance to look at 'Machine Models of Music: Minsky to Mozart' from the MIT Press yet? If so, can they give a bit of a review or minimally, a list of the papers presented? However a few people thought it was a collection of music! For the edification of all and sundry (From the MIT Press catalog): Machine Models of Music >From Minsky to Mozart edited by Stephan Schwanauer and David Levitt Machine Models of Music brings together representative models ranging from Mozart's "Musical Dice Game" to a classical article by Marvin Minksy and current research to illustrate the rich impact that artificial intelligence has had on the understanding and composition of traditional music and to demonstrate the ways in which music can push the boundaries of traditional AI research. Nov 1992 - 300pp. -50 illus. - US$39.95 ISBN 0-262-19319-1 Yes, my research is in machine perception of musical pitch and rhythm using connectionist methods. -- =--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--= | Leigh Smith _--_|\ Email: leigh@psychok.DIALix.oz.au | | / \ Tel (H): +61-9-450-2014 | | Perth --> *_.--._/ Perth, Western Australia | | v | =--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--= ------------------------------ End of the EMUSIC-L Digest ******************************