Hexells – Self-Organising System of Cells Drawing Textures [WebGL]
znah.net
znah.net
i found that the entire pattern changed rather quickly anytime i was enjoying it, which felt a bit too chaotic for me to spend more than a minute on it. cool page though
edit: seems like there is a socket syncing it.
I tried to disable it by entering 'window.demo.ws.close()' in the console, but the script just reopens the socket.
edit2: Ahh, if you "swipe" up it says "you control x screens" (x=126 in this case). Seems like there's always one controling it? Sometimes it outputs "control is busy". Weird thing.
Actually pretty impressive if you ask me
Also seems to be based on work mentioned here https://distill.pub/selforg/2021/textures/ which has a cool interactive widget at the top.
I haven't gone through it yet, but plan to in the next few days.
He posts/retweets pretty interesting stuff, such as https://nitter.net/evoluchico/status/1503307804043583488#m
I was re-watching one of them last night after maybe a 1.5 month gap and I was astounded how much better I understood it the second time! (The puzzle pieces seemed to fit together much better) Fascinating work!
For developers, here's a simpler but still cool GPU cellular automaton with code and a tutorial:
I wanted to add that I'm in awe in the very subtle way the colors are chosen, which has an almost Impressionist quality in some cases. Little variations of hue that are fairly neutral, but where you can see the rich color variations when focusing on the details. Sometimes there seems to be little patches of light in an abstract pattern, with a physical quality. The cell size can also create a perspective effect, particularly striking with a wave pattern. It's so evocative!
> that was trained to build textures using neighbour communication only.
How was it trained? on what data? I imagine trained on highly local patterns found in natural images, but I'm trying to imagine what the cost function might have been.
https://distill.pub/selforg/2021/textures/ - this is the paper the authors refer to. Here, the network is trained to maximize activation of another filter of a CNN pretrained network (VGG, for example), which typically correspond to certain textures and shapes. The losses are typical style/gram losses.
Given the music I/we make fits the trippy trance pattern quite well, this ought to work quite well. If I capture someone else also fiddling with the patterns, all the better. Perfect now it would be to integrate a lua bridge so I can remote control the pattern from a side-chain on the master bus in reaper ... if any of you has something like that figured out already, please point it out to me, thanks in advance!
(if any of y'all are in for some home-grown manual psychedelic trance, feel free to check out our music at https://www.youtube.com/channel/UCGDW9hicaeV9Z15gFpXcANw )
Please don't complain about tangential annoyances—things like article or website formats, name collisions, or back-button breakage. They're too common to be interesting.
Wolfram believes that Rule 30, and cellular automata in general, are the key to understanding how simple rules produce complex structures and behaviour in nature