Looking for your BIG DATA/generative tracks for my DJ set

The Lounge General Discussion
#1

hey there,

I’ll be djing on Monday after the opening talk of a new lecture series about A.I. at Favorit Bar, Munich, and it would be fun to play some tunes that were made with more than a little help from our friends: the MACHINES :smile:

So in case you already got your own expert systems endlessly permutating your sample libraries and synth engines, and are inclined to share, please leave a link.

I’m going to record and upload the mix.

cheers,
door

here’s a (sadly German only) link to the event:
http://www.kunstverein-muenchen.de/de/programm/veranstaltungen/2017/learning-to-walk-ki

#2

Hey, this is an old one of mine, actually the first time I tried something like this.
Feel free to add this one to your mix if you like.

2 Likes
#3

I wrote a mutation generator for a Lo-Key generative synthesis track using the ChucK programming language. Would probably benefit from some additional DJ play around the core progressions. Note, the bass on the track could be removed with a hi-pass filter most likely. I wrote the track in 2013, and if I recall right it is in C harmonic minor…

All sounds 100% generated programatically with ChucK’s Oscillators and Generators, including the kick and hi-hat.

The melody pattern was written out for the first 16 beats, and then a melodic mutation algorithm I wrote in ChucK continuously changed the melody throughout the song according to a pseudo-genetic set of rules I wrote – so the melody is almost entirely a one-time performance of my algorithm. A replay of this ‘song’ through ChucK would result in a different experience.

( chuck.cs.princeton.edu/ )

One thing I find interesting about this track, my sound card on some of my computers can’t play the lowest note (or two) of the bass track, which is at C1 at 32.7Hz (some of my headphones and speakers have the same problem…)

3 Likes
#4

Rad! I do a bunch of generative stuff, I’ve built a cool Reaktor environment with my home-brew sequencers and combining/mangling/routing schemes. I have 16 slots for building synths or sample players into and then a 16 track mixer and multifx section. Did a track combining it with the new OT beta OS yesterday. About half is from Reaktor and half from OT, but the whole thing is run into another instance of my self playing multifx box which also creates a lot of sonic diversity. I’ll pm you links to some other tracks that are 100% Reaktor generative grooves

https://soundcloud.com/knobgoblin/generativegroove-otadd3

Here’s another one that is from the same set-up, 100% synthesis and home-brew sequencing. some of the sounds are the same (mostly the drums and the physmod-ish synth) but the rules for the sequencing are different. didn’t have the multifx box working fully as intended in this one, so it’s use is a bit more limited

2 Likes
#5


made a new one for you today

2 Likes
#6

thanks a lot! I’m quite fond of these hesitant arabesque movements, strangley taking me back to Lut Gholein, also the piece fits perfectly with the sound of Fatima Qadiri.
Guess I’ll mix it with the two facebook chatbots babbling along :slight_smile:

#7

thanks man! lovely pacing and I like the way these glitch sounds complementing the pure tones tickle me.

#8

Thanks much! I just made the track downloadable…

1 Like
#9

yeeeaahh, that’s right up my alley! please tell me that’s available on vinyl :heart_eyes:

1 Like
#10

No vinyl unfortunately. I want to do a proper release of my generative stuff, haven’t gotten around to it though. SoonTM…

#11

they’re great :slight_smile:

1 Like
#12

Thanks guys! I know the shows already gone down but I thought I would share this here because I’m so excited by what my physical modeling synth is doing. Same beat as before, just changed the synth

https://soundcloud.com/knobgoblin/generativekarplussy4

2 Likes
#13

2 Likes
#14

even sicker now with that deamonic flute player! but I’m a sucker for Karplus Strong synthesis :slightly_smiling_face:

#15

hey folks,

ok, here’s the set, enjoy almost 3h of ambient post-dubstep noise:

https://www.mixcloud.com/daniel_door/ball-have-zero-to-me-to-me-to/

@musicmagus sorry that I put your track at the very end :smile:

thanks a lot for your contributions!

3 Likes
#16

hey folks,
looooks like that set was quite well received, so I’m…the resident A.I. lecture afterparty DJ now :alien:
the title of the lecture on monday is: “we cant’ feel data…” and in case anyone of you got a track using huge amounts of data or maybe even…highly personal data (browser history ??), please share and I’ll try to integrate it into the next set.

going to record and upload the mix, again :slight_smile:

cheers, door

old(ish), but (conceptually) gold(ish): 70200 samples GEMA fuck up by Johannes Kreidler:

#17

Ive never attempted generative stuff, really interesting music man. Spiked my interest a bit, Ive got limited reaktor experience but never really delved beyond blocks and prebuilt stuff. Care to explain a bit further how you go about how this works or point me in a direction I can learn more? Also an octatrack user so I’d like to have a go at something similar

Also @door , how is this data being translated into audio? I’m curious and struggling to find much on it

#18

I just started a project to do this via feeding a deep learning system a bunch of dissassemled midi from 20 years of a friend’s work, but I am still super early in the doings. Haven’t decided many things… Tensorflow or Torch? And then how to turn the midi into meaningful bits (probably after normalization into a specific key) to train the neural nets. I suspect I’ll be taking a few years as this is just in my spare time.

#19

hey @jb, I’m sorry to say I’m just a noise musician and DJ, not a developer, but
I’d say it basically comes down to feeding an algorithm huge amounts of data (music) and let it sort out similarities and discrepancies (and define a goal).

have a look at this cdm article - since listeners seem to tend to skip to the drop of an …edm… track (Y THO???), an action that is logged, generates data, this app can predict drops in tracks it hasn’t been fed with yet.

or, as the project mentioned by @sigint, you can train a neural network on those huge amounts of data with the goal of it making sense of it and producing something similar. here’s a cute video showing the basic idea. aaaaand here’s a bunch of rock’n roll coders of a not so cute tech giant showing this can be done in real time now…

with the OT, well, OS 1.30B didn’t make it self aware (yet!), but trig conditions are a set of rules, that influence the way of ones human-machine interaction :slight_smile: