dear forum.

i'm looking for a convenient way to create simple melodies and lines out of a short hex string. currently, i'm working on a geeky text-based drum sequencer in pdlua which is already capable of creating creating a binary drum pattern from simple text/hex strings. now i want to add a melodic dimension for synth blips or basslines which are somewhat expressive, but relatively simple to construct and type in.

so far i have come up with this 4-nibble (hex digit) version:
"xyzz" where x is an arpeggiator function, y is the arp function parameter and zz is the base note. the arpeggiator functions might be mapped like this: 0 = ramp up 3 notes then repeat, 1 = ramp down 3 notes, 2 = ramp up 4 notes, ..., f = random note...; the arp function parameter is most likely the interval that ramp functions jump up or down. the result of the arp sequence gets added to the basenote and the resulting melody is then mapped onto the already existing drum beat structure that is sequenced independently.

do you think this might be a worthwhile approach, or am i wasting bits and overlooking some obvious solution? maybe a array of 255 preset melodies to choose from, so 2 hex chars for this and maybe 1 char as an transposition argument or for reversing / inverting the melody or...

i'm not looking for a way to store an arbitrary melody, but for a way to create an expressive line out of as few chars as possible. i'd be interested to hear if any of you had other ideas, or maybe even ideas of arp functions fitting into the above model.

yours truly,
syntax_tn

I've been contemplating doing something similar with Lsystems for some time (http://en.wikipedia.org/wiki/L-system). However my brain has never been up to the task. an Lsystem patch could build some pretty neato melodies, I bet.

Not to divert this thread, it may be deserving of its own, but thanks for introducing me to l systems sunji.
Now that i know what they are called i can learn more about how they work.

I can see them working in visuals, but how would you go about making this work with audio?

edit: ok, i'm starting a new thread.

:You are taking a very 'synth' (arpeggiators, linear harmony, etc) angle on composition here, not to say that is inferior. In fact, I think, from your current standpoint, you have a pretty good start. One thing I would suggest, as a student of serialism is that you check out Arnold Schoenberg's techniques for serial composition, especially matrices. Included in such a matrix is a (non repeating) tone row, as well as all of it's possible transpositions and retrogrades. In his system, a simple 'data set' implies much much more than just a melody. You have already experimented somewhat with serialized rhythm with your drum machine, you could apply similar rhythmic patterns to a tone row, and then have some other nibble affecting transposition or retrograde/inversion.

If your interested, or if all this lexicon is above your head music theory wise, let me know and I'll post (or patch) to explain further.

hi.

i'm familiar with the basics about schönberg's twelve tone technique, but serial composition does not ring a bell to me.
anyway I'm not sure the result would fit into my current concept. actually i'm not looking for anything that goes too deep into composition but rather for a lightweight accompaniment to the drums (which might later grow into more).
for now, i'm experimenting with a simple variation of the thue-morse sequence which is able to produce self-similar melodies but lacks variation after a while.
I definitely need to read up on serial composition, it might be a nice addition to my palette.

another thing: you accused me of taking a very 'synth' angle on composition. i'm afraid this is even more true as i'm thinking within a rastered box of time signatures and note values (even if in a different syntax than traditionally used). i should expand my horizon to synthesis approaches that go beyond "play note X at time Y" and play with sound synthesis in a somewhat 'freer' way (e.g. evolving pads, noodling modular synths), and then find a suitable 'notation' for this. but for now i should rather focus on the basic drum sequencing and user interface.

thanks so far everyone.
syntax_tn

the whole 2nd viennese school is worth looking at, Berg and Webern. Webern took the serialism to it's limit. Also Arvo P"art has an awesome systemic approach to composition (anyone got a name for that?). Brian Eno has a bunch of machine composition techniques that I love (most of them are based on degrading tape loops).

I'd take a look at Per Noergaard's work on melodies derrived from his Infinity Series work. That may turn out to be very useful in the context, and much more pleasing to the ear than the second viennese school/serialist/12-tone stuff.

ok, i'm reading up on noergaard and i just read that under certain conditions, the infinity series can be a thue-morse.series (which i'm already toying with). the thue-morse-sequence can itself be created by an l-system as sunji recommended in the first reply. the infinity series can be built (or rather looked up at playtime) very simple, too, by bitcounting (which i already implemented for thue-morse). so things start falling into shape here. now i need to represent sets of parameters for these methods, textually.

thanks again.
syntax_tn

Just to throw a random spanner into the works

State machines can be programmed very compactly, which is the idea behind 'Wolfram tones'.

some simple harmonic arpeggiator:

step
t f f
tabread seq
mtof
mod 8
* 3
mod 8

• 1
  *~
  osc~

hi.

ClaudiusMaximus: nice one. it also works when the tabread seq is altered by a variable [mod x]->

• -> [whatever z] -> ... combination, that way i can define nice basslines by just a few variables. by the way, thanks for pdlua, the logic of my drumcomputer would not be possible without it.

syntax_tn

ok, here's alpha alpha proof of concept, more noise than music, apologies for the dropped frames (my rt kernel is on the fritz).

what I've got here is a random-voice-leading generator that accepts a beginning pitch and an acceptable range before reset (lo/hi), eg,

|440 220 880(

will start the patch on a440 and randomly select a tone enharmonic to it's current pitch every time you send a bang. if the tone falls below 220hz, then the harmonystring will be biased to pick higher notes, and contrarywise it will be encouraged to move back down if the pitch rises above 880.

I have 4 of these harmonystrings controlled feedbacked with themselves and against eachother, and then using maxlib chord detection I have a sine tone acting (sorta) as a walking bass. I also use some maxlib beat detection routed to Obi's efficient drum machine, and voila! I now have an alcoholic cat jazz trio synthesis.

let's try a shorter sample, for file size.

http://www.pdpatchrepo.info/hurleur/youaresmallenuf.zip

@sunji said:

the whole 2nd viennese school is worth looking at, Berg and Webern. Webern took the serialism to it's limit. Also Arvo P"art has an awesome systemic approach to composition (anyone got a name for that?). Brian Eno has a bunch of machine composition techniques that I love (most of them are based on degrading tape loops).

The Arvo Part compositional technique you are asking about is called Tintinabulli.