I want to build a midi perfomance setup

[disturb]

Hi folks, first post here eventhough i've been checking out the board for quite a while

This is a pretty important question for me, as i feel it will be quite an important part of my musical evolution, so thanks in advance to anyone who'll take the time to read and comment.


Basically, i can't play the keyboard, and as far as composing goes, i feel quite restrained by the whole piano roll paradigm of 99% of Daws today.
So I've been meaning to build my own composition setup that would allow me to do things in a different way, my way i guess...
I've looked into Max, Chuck, SC etc, but I'm really unsure about which option will be best suited for what i want to achieve.

I'm gonna try and explain what my goals are, and i hope i'll get some feedback from users of different softwares.

I'd like this 'composition canvas' to be a modular environment, so that i could build and reuse all kinds of modules for arpegiation, chords etc, and define different paths and setups for different purposes.
It would be algorythmic in a sense i guess, but not based on pure randomness, the emphasis would be on user input ; wether it is from a controller, programmed note/gate sequences, or live playing/triggering of modulations and whatnot. So to speak, it would allow me to build my own stairway to go from one point to another, putting me in control of every step in the whole process.
Interaction would be the key, i'm not sure i'd be satisfied with anything that can't run live and be played in a sense.
My goal ultimately is to be able to build something that will be musically meaningful TO ME, and that i can play and interact with.

That probably sounds a little too abstract (and it still is for me tbh), but here is a practical example of something i'd like to achieve.

A sequencer to build chord progressions :
-it would work in scale degrees
-could easiely be switched to different scales
-would allow you to sequence by directly inputing a degree/chord type/inversion
-would somehow be 'smart' about the process of fitting chords to a scale, rather than blindly force an inappropriate chord to a key (maybe dynamically rearrange the chords list for each degree, according to the degree of pertinence with the current scale perhaps)
-have the capabilty to do such things as automatically select inversions to bias the voice leading. For example a 'realistic' mode, that would pick the inversions, so that the keys are as close to each other as possible ; pretty much like a keyboardist (maybe not a real talented one ) would do.
The control over how the voice leading evolves is a good example of the type of interaction i'm aiming at.



I really hope that made sense, and that some users of Max, chuck, athenaCL, SC, CSound... will be able to comment on wether what they use would be a good way to achieve this.
That will be a long and hard process anyway, so i'm really sweating over a false start...

Cheers !

[kijjaz]

wow!! i'm really excited by your goals now..
you've got such a big and exciting start.. and much bigger than mine now hehe.
so it's really nice.
and i'm looking forward to your ideas ^_^.. in the near future i hope!

by the different features you want for the system..
i guess there are more than 2-3 softwares you can use..
and each might serve better in a particular feature.
i'm not very good in all softwares / languages..
but i'll try to recommend from the software i've played with.

Csound, ChucK, Pure Data or Max/MSP

from most of your requirement, i recommend using Pure Data / Max/MSP
mainly because it'd be convenient to access the interface..

but for me i also recommend ChucK
sometimes i can't program some crazy ideas on puredata..
so i need to get into a deeper style of programming.
now chuck has more ways to build interface (on Mac OS)
so for a very algorithmic ideas, i recommend chuck.

but i was very in love with Csound 2-3 years ago in the flexibility i can do in programming..
and there are lots of Ugens to play with also.
but i moved to program mainly in chuck because it's easier to program for real-time performance,
and tweaking with the code while performing live is much easier -- the code is easier to read.
- - -

now, most of my patches for live performance are on PD and ChucK
and they're going well. i hope i can hear more from other artists also..
thanks in advance also for later ideas for this issue ^_^

[v-un-v]

[disturb]

[disturb]

[blue hell]

Maybe have a look at http://electro-music.com/forum/topic-7812.html and then maybe at http://electro-music.com/forum/topic-10436.html for some stuff on this forum that I got reminded on by this subject.

[disturb]

[disturb]

I'll just quote mosc from one of the two aforementionned threads :

[blue hell]

Are you going to write a program yourself ? I assumed this from your first post, but after re-reading I'm not sure if you really wrote that.

Anyway, the user interface, that is how you would like to control your ideas, would be an imporatnt aspect I think. Howard' s ideas are still pretty abstract (in not directly suggesting a user interface), but my hope was that the threads I mentioned aboce might help you to get a better idea of the interface.

I'd not be much use helping you there I guess, as for me Tom's (v-un-v) idea of using the Nord Modular has become the preferred way. That however seems to go a more abstract road than you have in mind.

Maybe have a look at KeyKit .. ?

BTW on the forum.

[bachus]

I've never used keykit but from what I recall of reading some of the docs it might serve your purposes. It's free and the guy who wrote it sometimes shows up here. Anybody else want to comment on it?

[disturb]

Hmm well i'm totally unexperienced when it comes to coding, so i'm not even thinking about doing this in C++ and he likes...

What i'm trying to do with this thread is get some insight from users of different music-oriented development platforms (probably not the best choice of words), to figure out which software is worthiest for this purpose.

The main flaw here is probably the lack of specifics regarding the inner workings, but as i have no previous experience with such things, i can only try to put the global concept into words and hope it makes sense for someone...


Also i've had a look at keykit sometimes ago, but was a little put off by the lack of ressources. And unless i missed it (see previous sentence) i couldn't figure out a way to use different modules in a user definable path (modular). I think data can only pass from a module to another through copy/paste type functions, which would probably prevent me from implementing the setup the way i envision it.
But tbh i sticked to the included modules and didn't have a good look at what's possible with the language.
I'll have another look at it.

[bachus]

[electroboz]

if i may suggest. you should think about having a cheap midi keyboard. even if you cant play. you don't need lessons you can just noodle around. the more you play the keys the easier it becomes. it is by far easier to press some keys, hear a motive you like, and hit record then it is to put notes into a piano roll by hand. just a suggestion. my first keyboard was just a 2 octave controller. it works fine for coming up with a bass or lead line but i did grow out of it. its suprisingly easy to come up with great sounding stuff by recording into a 16 note looping sequencer.
just my 2 cents.

[disturb]

@bachus : ok i'll have a better look at it, i'm warry with keykit i might have to do it all into one big module, as opposed to several specialized building blocks interconnected ; which seems to me less flexible and harder to do. I'm also unsure about the possibilities to actually control part of the application with midi, i'll check it out...

@electroboz : i do have a keyboard and fiddle with it quite often, it's just that i'm far from capable to play what i want.
But in this project, the keyboard would be used a lot ! Wether it is to capture a riff to tweak, or some sets of intervals for modifier modules, a simple rythm for a gate sequence, chords etc...

[Digiton]

[Digiton]

OH and disturb, check this shit out.......
http://www.obsolete.com/120_years/machines/free_music_machine/index.html

he is known mostly for VERY tonal (but beautiful) music, alot of "remixing" old irish folk tunes, but he had another bad ass side to him that due to technological restraints of the time could never be realized.

[disturb]

thx for the kind word digiton, and yes i agree with you to some extent on the tight relationship of composition in a traditionnal sense and sound design.
Both benefit from the other, but i have yet to find out in which way my sound design will evolve with better composing skills...
I'm thinking of intertwining different sounds for each voice of a chord, and dynamically map the notes according to the voice leading and things like that. Sounds like fun...


Anyway, check this out guys : http://www.mloco.com/thesis/
Fun little thing, i like the idea of the visual feedback, i'll kep this in mind.

As an update on my project, i've been reading up a bit on algorythmic composition etc, i believe MAX is my best bet now.
Most of the papers i've found are implemented in MAX, and that will help getting started.

I'm starting to have a clearer sight on how it will work.
At the heart of the patch there will be a chord sequencer, a bit like the link above, but more flexible of course.
I'm thinking fuzzy logic will be a good way to get started on the mechanics of chord analysis etc.
I need to read up on colls, to find a way to dynamically display different sets of chords depending on the selected scale, and their level of consonance/dissonance.
Then it's gonna be all about building little modules that will use the data from the core of the patch, and alter the rythm, modulate pitches, arpegiate etc...

I'll look into key modulation at some point too, but i'll need to dig in music theory first !

Let me tell you that i'm way in over my head here, both on a theoritical and technical level (hey at least i'm honest :p) ; that will take me a lot of time i believe, and i'd be very glad if anyone would like to take part or help out in any way.

cheers


PS : This post probably is a little more specific about what this setup is about, so if anyone has some bright idea about a software that'd be more suited than max/mxp, speak now, cause i'll soon get started

[Digiton]

in terms of COMPOSITION, i believe "your ear>music theory" if you catch my drift. not to stop you or anything, but its just my opinion. If your concerned about formal classical stuff, get some mozart mp3's cds, he rocks when it comes to how "classical" music is structured. Venetians snares rhythms? no. Stravinsky harmony? no. But its formal construction is of the hook! and beautiful too.
not to spam, but these two songs i recently posted (as with all my recent work) deals a lot with abstract sound/design and its relation/contrast to tonality (although sometimes I just like to get noisy ). It might not be what your looking to do, but here is the link.
http://www.angelfire.com/planet/aveles/music/mf.mp3
http://www.angelfire.com/planet/aveles/music/sound.mp3
OKAY I'm outta here! good luck

[mosc]

Bachus alerted me to check out this thread.

Keykit is pretty much midi only, but it is IMHO the strongest language for algorithmic types of things because of the note/phrase structure quoted by disturb above. Keykit supports that exact concept.

In some ways, it's a great idea to separate the composition component from the sound generation components, but there are pros and cons to each approach.

Keykit is very modular. You can write functions to generate notes and phrases and pass those to other functions - much like one would do with patch cords. It is different from MAX/PD mainly because keykit uses code and MAX/PD is based on the graphical interface. Personally, I prefer code.

The keykit language is superb syntactically, again IMHO. For music, I would much much prefer writing in Keykit than C++. (Keykit isn't suitable for writing operating systems though.)

Your goals are appropriate but I think you'll have to spend some time looking at several alternatives before you find something that you will find something with which you will be most comfortable. Don't overlook ArtWonk.

[disturb]

[dewdrop_world]

Hi disturb,

This sounds a lot like what I am doing in SuperCollider. I'm not quite there -- I have yet to tackle the problem of handing control from live MIDI note input to algorithmic sequencing seamlessly (with the key word being "seamlessly" -- I can get the MIDI notes in in real time, and I can feed them into algorithmic processes, but doing it without a perceptible break is something I haven't figured out yet).

You might be interested in some of the elements of my framework:

• ModalSpec -- Converts MIDI note numbers into a modal representation, preserving accidentals. You map the notes onto the mode for storage and processing, then map them back onto note numbers at play time. If you're using open sound control for note triggering (that is, with SuperCollider's native synth or another OSC-compatible sound module), I have some companion classes for non-ET tuning.
  
  
• MIDIRecBuf -- A container for note data. Includes built-in methods for quantizing the notes, and parsing them to identify chord notes and grace notes. Grace notes, for instance, get "attached" to the following main note, which has a whole bunch of advantages. When you're manipulating the note data algorithmically, you don't have to worry about separating a grace note from its main note and thereby getting unwanted results. Also, when quantizing, the grace notes won't mess up the calculation. The main note gets quantized, and the grace note is adjusted to fit in proportionally.
  
  Parsed notes know how to stream themselves out in order, so it's very easy to sequence with them after processing.
  
  
• I have several prewritten "process prototypes" that hide the really complicated flow of control from the user, and expose parameters that you can use to plug in your own note processing algorithms. I get a lot of mileage out of them. The problem right now is that I haven't had time to document them well, so you might have to ask me a lot of questions by e-mail which I might not always be able to answer quickly.
  
  But I think it has a decent amount of power. I didn't want to have to rewrite an algorithmic sequencer from scratch for every line of a texture -- I wanted to write little code bits for reusable, customizable algorithms and then reference them by name in the process parameters. (That is, modularity and reusability were TOP design considerations.)
  
  For just an example of the kind of power, the choice of algorithm on each iteration can even depend on the note data. So, in one piece I added up the note durations and if they were greater than a threshold, I would weight the choice of algorithm towards functions that would delete notes (make the phrase shorter); otherwise it would tend to expand the phrase. The result was a line that breathed organically on the macro scale, gradually inflating and deflating. I was delighted when I realized what it was doing, because the code I had to write outside the process prototype was actually pretty simple.
  
  
• Also important is that the note output mechanism is abstracted away from the sequencing code, so you can choose which output device to use with only minimal changes. I use scsynth, the native audio engine, exclusively, but you can also use MIDI output, for example. (That's not unique to my library; it's a feature of the pattern-stream protocol in general. "Pattern" in this sense does not mean an n-step sequencer; it's much more powerful.)

Is that enough of a sales pitch? Oh yes, and it is all geared toward realtime use.

James

[mosc]

[disturb]

@ James : WOW, this sounds mighty impressive ! Seems like we're on the same wavelength.
Your sales pitch really made its effect indeed, your design takes into account pretty much all the things that are crucial to me ; i really liked the part about the modularity, and the way you handle grace notes in the midi buffer seems really clever...
Too bad, that's another option to seriously consider


I have a few questions for you, if you don't mind :

-Your website seems like a very cool SC resource too, I was wondering if the patch you talked about is the one available there, tarball, since you seemed to imply i'd be able to check it out, but i'm not sure it is what you're refering to...?
-When you mention your problem with what seems like latency with midi control, is it because of the too heavy computational problem, or something more SC specific ? Live tweakability is really crucial for me.
-Is your framework geared towards tonality (seems so, but i'd rather make sure)? Is tonality something that's easy to handle in SC ?
-Is SC really a stable platform on windows system (eventhough you seem to be a mac user)? I seem to recall people mentionning a lot of instabilities on xp, maybe that's been fixed.
-And finally, did you dvelop this on SC because you were already using it, or did you purposefuly chose it for this framework ? If you had to make a choice between SC and another application, what tilted the balance in favor of SC ?


@ mosc : hell this seems pretty convinient. And i'm unsure if code actually bothers me or not, i'll have to learn something new from scratch anyway...

[dewdrop_world]

It is available on my web site. There are two packages: dewdrop_lib, which contains more fundamental support classes, and something I currently call "chucklib" but whose name I will probably change at some point (because you might get the idea from the name that I'm trying to duplicate some of the structures of ChucK, but that isn't the case).

Documentation for dewdrop_lib is in pretty good shape; for chucklib, it's very spotty. One of the files has some examples, but they just scratch the surface. Most of the features are not documented yet.

There are a couple of problems remaining with live MIDI input. First, I had it working early on but then changed some structures internally in the process prototypes. That part of it is still buggy (but this conversation motivates me to dig in and fix it). Second is the problem I mentioned about handing control over seamlessly. The problem there is not computational requirements -- all the sequencing stuff is very cpu-cheap. Rather, it's that to ensure proper timing, sequencing routines in SuperCollider should execute slightly ahead of the time they're supposed to sound. Messages to scsynth can have a timestamp slightly in the future, so that small variations in the time it takes to transmit the message can be corrected. But, live MIDI input should have no timestamp so that the sound occurs as quickly after receiving the MIDI message as possible.

It's a time travel problem -- if you stop MIDI input precisely at beat 100.0, but your sequencing routine should have started at beat 95.5 so that it can make sound exactly at 100.0, how do you travel back in time to start the routine? Don't have a good answer for that yet... but talking through it here gives me a couple of promising ideas. Probably messing around with the timestamps would do it.

SC has built-in methods like keyToDegree and degreeToKey which convert between scale degrees and midi key numbers. I wanted more flexibility and a consistent way to store scale definitions, hence ModalSpec. Short answer, yes, you can do a lot with traditional or novel scales in SC.

Windows... still under development. You can get started learning the language but I can't consider it quite ready for prime time. It's not unstable in the sense of crashing a lot, but the scheduler chews up cpu like a dog who got into your sneakers -- everything is perfectly responsive but task manager shows your cpu pegged. I'm also not sure if MIDI is supported in Windows yet

You might be better off setting up a Linux partition on your HD. The word "Linux" scares a lot of people but actually it's very pretty. (I sometimes joke about Linux but honestly, if OSX ever ceased to be a viable option [the Woz forbid], I would go to Linux sooner than Windows without a second thought.)

I chose SC the second I saw how powerful and efficient it is. I kind of stumbled forward at first with some structures that were not very powerful, and after a near failure in a performance... three years ago, gosh, doesn't seem that long... I reassessed my needs and rebuilt how I wanted to do sequencing from the ground up. That investment has paid off many times over.

James

[disturb]

Thanks for the answer James, still processing all this info

BTW, i just found out that IRCAM's OpenMusic is open source, seems like the distribution fo linux has started, and windows seems to be on its way, anyone ever tried this program ?
Looks like everything i could wish for...