now in the user library: vocode for freeeeeeeeee

made this vocoder with core NLB ladder and butterworth filters, and since it sounds dang good, figured i'd upload in commemoration of daft punk's closing the final chapter on their long and storied career

serious vocoder | Entry | Reaktor User Library (native-instruments.com)

 

Thanks, Andrew! Tell us something about how you went about this?

 

Mayhaps, french Daft Punk in going home?

 

Demo sounds good.

I made one a while back using the library ladder BP filters, and was pleased with the results.
For extra mojo, plug equinoxe deluxe into it - get that divide down strings vibe... like many of the classic vocoding sounds.

 

the onl

aside from the choice of filter, the only thing non standard about it is the that envelope folllower for each band uses something between the amplitude and the square of the amplitude, which seemed to tighten up the sound

actually the most interesting part was making the unvoiced excitation envelope:. three envelope detectors, one of the full modulator signal, and the other two of bandlimited portions of it (>500k and <2000k). the bandlimited envelopes are divided by the main one. whatever is left of the hf envelope is multiplied by one minus whatever is left of the low one, so that the noise will be loudest when there is significant high frequency energy (relative to the total energy) and less below 500hz

 

of course you're right, and now that you mention it i feel a bit silly at having not even considered putting one in. every vocoder needs an internal divide down cheesy string section

 

My dude...everything benefits from more cheesy divide-down string machines.

EVERYTHING.

 

I've been building all my vocoders with the primary 4p bp filters , these are good enough ..easily 512 bands etc ( voice module is a must )
Fun starts when replacing the carrier filters with sine wave osc's instead , to reconstruct it
Up to 512 partials for an almost perfect reconstruction , ofsetting the frequencys for the modulator filters for formant shitft etc ...

 

oh yes, i've done big vocoders. have been fully down that rabbit hole. yeah, the primary filters are fine, especially when you want 500 of something. are you just raw dogging 512 bands, or have you doing signal analysis/ transmutation stuff? i started out modding sparky from the user library, stripping away its design flaws one by one.until what I ended up with was capable of some fairly incredible things, sadly it is long gone to data loss

here's a tip: find all the frequencies that are louder in amplitude than their two closest neighbors. these will be harmonic peaks, and you can take the amplitude and the amplitudes of the neighboring bins or voices and use this formula to find the peak's fractional value, which you can use to tune your oscillator towards the actual frequency. not only will you find that you don't need quite as many voices, but having 'peaks' and 'not peaks' separated out also opens up a bunch of new possibilities

at one point i had sparky modded as a spectral sampler, but using gabriel mulzer's method of defining ones own bins and getting amplitudes for sine cosine pairs, working out amplitude/frequency. with this i was able to get down to about a hundred voices/bins, and while it wasn't a perfect reconstruction (pretty much, you need fft for that) it was full fidelity and everything tonally important was in tune. though i was never able to get the phase to frequency to work correctly again after trying to rebuild it from memory. nothing about it was terribly complicated

i think my favorite discovery of all time from that era was that i could record the same amount of filters' output envelopes to event steams at about half CPU by running all the event processes from the display clock LOL

edit: whoops forgot to include the picture

 

actually i was completely grossed out by a polymoog experience once, a video of a church keyboardist playing his 'guy in a flashy band' licks up and down the keyboard .incited me to write i could think of, something to the effect of 'stick to your yamaha stage piano buddy, that's not how you play synthesizers'

was thinking probably the reason divide down works so good as fake strings might be because the sound with each tone being phase locked is kinda reminiscent of the mode locking that occurs when using a bow to excite a string?

okay just parsed that everything in bold red with winky face. i say polymoog and vocoder carrier signals can stay, but if there was a button to huck every last electric organ into the void, i wouldn't even stop to think about it

 

as long as they stay there

 

also, a neat thing you can do with all those sine oscillators, is i think what is called raised cosine pulse train synthesis, you take a master sine wave that you want for the pitch, as well as its corresponding cosine with DC added so that its lowest point just hugs 0. then you multiply all those sine waves by the raised cosine, which perfectly windows them for the original sine, and then i think you just amplitude modulate it, if i remember correctly, something like that. that was the best sound i ever got out of a big vocoder