1,974 karma · joined March 9, 2007
https://www.a1k0n.net/
The neural network is trained to mimic collaborative filtering vectors from raw audio. It's a separate model from time signature, key, mode, tempo, loudness, etc.
of course the same is true of ground stations... I'm not actually sure how they do it but they probably don't need as high-gain of an antenna to reach them.
xadd eax, edx
loop -3
implements the Fibonacci series in two machine instructions (notice eax and edx have 514229, 832040 in them after you run the example, the 29th and 30th Fibonacci numbers). I first saw that in a book a friend got me for christmas; a hard copy of this site: https://www.xorpd.net/pages/xchg_rax/snip_00.htmledit: nevermind, I thought you could specify a startTime in addition to endTime but you can't. so you'd have to use audioContext.currentTime to figure out where the sweep should be now, ramp to there, and exponential sweep down to the final value at the end time.
It still needs accent emulation which is a bit of work, high pass output filter (it's bassier than a real 303 because there's some capacitors between the filter output and the amplifier I'm not emulating), the knob ranges / nonlinearities aren't quite right...
edit: And it also needs the square wave, which is super weird on the 303, as you're aware (yours has a sample for it I think)
Basically, not at all like a 303 except in spirit
https://github.com/a1k0n/gsynth/blob/master/modules/tb303/tb...
It doesn't sound very much like a real tb303 whatsoever. I'm working on a JavaScript one based on an x0xb0x I built, but it has a long way to go.
The envmod knob also affects both the max and min cutoff values from the main envelope generator.. that part is tricky - the envmod and cutoff knobs are exponentially tapered, but the filter control circuit also contains an exponentiator so the sweep is sort of doubly exponential but its hard to tell the difference.
If the aperture is smaller than the jump height, then you need to do something smart to time your jumps.
Sadly, that was my most upvoted comment in ~9.5 years on this site.
What optimal parameters did it find for these?
https://github.com/golang/go/blob/fad2bbdc6a686a20174d2e73cf...