hey friends. today i made a really rough fft. the first attempt crashed my laptop and deleted itself, which i suppose was a good thing. as far as i know (??) nobody has tried this on AT before. though i would love to be wrong about this!

edit: people have done vocoders before— 27-band vocoder in audiotool and i'm not sure if this is much different or not? idk

if this kind of thing ends up being functional, there's some potential for cool frequency domain effects such as: pitch shifters, pitch inverters, spectral holds/custom reverbs, etc.

the deets;

64 bins at 1/7 octave intervals (from 20hz-20khz). the resynthesis is with very tightly bandpassed pink noise, each controlled via reverse sidechain.

the samples heard are, in order; that "yeah" sample, computerized countdown, some drums, morgan freeman talking about something, and a sine sweep from 20hz to 20khz.

things i'd like to fix:

- use a different filter to avoid both the slope's 10k limit and the lack of bandwidth control

- find better settings for the sidechains to create a more natural-sounding decay

- optimize processing size so i can do x128 or even x512 and up (!!??)

- use smaller bins or find a different algorithm that sounds more harmonious/natural

if anyone knows a better way to do this please let me know!!

Create an account or to write a comment.

  • if that's vocoder in audiotool, i'm never making a vocoder

  • that snapshot is terrifying

  • Awesome! Now I can use my laptop both as a vocoder and a frying pan. :P Seriously, nice work! :)

    • You're welcome. I thought I pressed the follow button last week, but for some reason beyond my comprehension, apparently not, so here it is. :)

    • i sat chuckling at this for a solid few minutes. thank you frigolito! honored by the follow as well :))

    • I don't know the technical things either, but good luck :)

    • ideally all the bands will be so tight that it would sound like the original sample. i'm not really aiming to use this with voice/pitched material but instead for resynthesizing and manipulating sounds. would that still be called a vocoder?

    • hey potasmic! i suppose in execution, it's quite similar to a vocoder, but the desired end result is pretty different. i'm pretty new to dsp and all this technical stuff so please forgive me if i misspeak!

      i guess it's not an fft because i'm not actually doing a fourier transform, but i am hoping to effectively translate a sound into the frequency domain where i can then do spectral manipulation stuff.

  • That project is massive lmao

  • imagine listening to this without context, would be creepy as hell

  • That's so funny, I tried to use the Morgan Freeman sample for a vocoder, but the response wasn't quick enough to make the high bands work. 60 bands is impressive for a voder, especially on AT. How did you divide the bands up?

    • hey noir! if you have a link i'd love to see what you did :) i divided up the bands equally 7-per octave from 20hz to 10khz.

  • There's another one but I can't find it rn

  • "10,9,8,7,6,5,4,3,2,1..."

    *plays thomas and friends*

  • this is so cool

    • oooh yes hopefully this can be used in some musical context that would be dope! also i just repubbed it with vocal samples bc i remembered this jonathan harvey piece (speakings for large orchestra and electronics) that makes an orchestra sound like speech. there's lots of potential for fun here i think

    • u could pull of some sick shit if u resampled this

  • Republished

    better demo samples (speech)