lifter

Official support for: xoxos
Post Reply New Topic
RELATED
PRODUCTS

Post

https://www.kvraudio.com/product/lifter-by-xoxos

scene: its last weekend, i've just finished kaze.

and just like every thing i've developed for the last month, as soon as i finish it, new idea.

kaze made me happy, it has emulative utility (all those years of hearing julius do it and no one in vst is doing it) but those fixed partials were kind of a bummer. but i think:

lets print two sets of cepstral impulse, 90 degrees out of phase. then frequency shift them to achieve a compressed or stretched series of partials, which will make it much better for eg. percussion.

of course the fundamental of the oscillator defeated a purely frequency shifted output, so that each partial has two sidebands. very well established synthesis terrain, and applicable for emulative task.

its a better platform now, inharmonicity mak it gud. still "not your rich guys cepstral resynthesis" since the source partials don't make it into this paradigm, but this can do mad audio holmes.

using a complex form also allowed for facile stereo output. of course, this sounds like its rotating (more or less with some sideband configs) but that can be lovely, and slight modulation can obscure the cyclicity if that not good.

fiddled with some of the inherited-from-kaze stuff a bit, this will load non-44.1kHz files better, some of the time? dunno, life doesn't give me infinite time for testing.

"next" it would be nice to have a more advanced noise specification (eg. a level) but tbh improvement here is likely to be time consuming/back to serious r&d and not on my schedule. i seriously didn't want to spend a month coding synths just now.. cheerful fun for old timer(s).
you come and go, you come and go. amitabha neither a follower nor a leader be tagore "where roads are made i lose my way" where there is certainty, consideration is absent.

Post

oh: important note i forgot...

due to the sidebands being produced by complex rotation, setting "sideband frequency" back to zero is like setting two oscillators to the same pitch after they have been detuned.. creating an arbitrary phase relation, which will have cancellation and might produce drastically different tones.

"gate sync" will reset this phase relation on each note trigger.

(personally, a param to specify the relation on reset would have been nice for me, but imagine trying to explain that.. so you get one complex rotation phase reset position..)
you come and go, you come and go. amitabha neither a follower nor a leader be tagore "where roads are made i lose my way" where there is certainty, consideration is absent.

Post

post 3 = file update soon

while i suppose much of the theory packaged here i am familiar with, i have had only so much authentic time with either build. from what i did experience, i think it has to have a phase param and would be frustrating otherwise. bbl.

this does mean patches saved in current release will not be compatible, seems to be that way when adding params.
you come and go, you come and go. amitabha neither a follower nor a leader be tagore "where roads are made i lose my way" where there is certainty, consideration is absent.

Post

done :)
liftervst.zip is current

adding the extra param did *not* cause issues reloading my demo patches in my daw. sideband phase isn't a mod destination (that takes some doing with my structures), but does respond to use/automation and is set on gate sync. not that different from modulating sideband frequency anyway.

there is some modulation of harmonics amplitude with it, along with the perceptual l-r thing possibly. i don't quite see how that happens as the spectrum should remain flat, but i've not used impulses in quadrature before.. i anticipate some ripples with the fundie.

out of bud except for work, no more dsp for a wile.
you come and go, you come and go. amitabha neither a follower nor a leader be tagore "where roads are made i lose my way" where there is certainty, consideration is absent.

Post

it is unlikely that i'll be touching this for a minute, i usually don't mess with audio unless its the right time and that is weed time. so if you want to play user input with the old 32 bit guy, here are some options i'd be inclined to change if pressed.

i guess kaze runs 3-6% on my single core, lifter 5-9% depending on parameterisation. that is about as high as i would go for a composition environment vst (btw graphically automating read position would allow syncing of time features to bpm for instance).

.....128 * 16384 samples used for noise aren't very long at 44.1k, and could be terrible if trying to do slow or static read position with unvoiced signal.. increasing the length of this array would be trivial. it would still be a problem. perhaps the best solution would be, render two sets of tables for noise, crossfade between them at some disharmonic rate.. one more linear interpolation, set of array reads, and phase operation... maybe 1% cpu to add...

..that won't change gui or much adversely, i think that's worth doing..

....the scale button expand and compress options are most suited for the 2 band modes, because they most resemble frequency shifting. this button could be removed, and the function baked into more options on the 'band' selector.. which at least offers a gui indication. lazy me. but it would eliminate stretch/compress options with 3,4,5 bands.. leave or change?

other options.. 1 voice poly because other wise it takes (voices * time) to load.. which can get serious.. (never got shared tables to work in se sdk) .. i could swap the standard midi to cv module for mine, which only works in 1 voice poly, i think eg. 'out vst' uses it.. it adds immediate/unsmoothed velocity response, nice for plucks.. and tuning/temperament options.. left it alone bcs easy..

..should i buck up and just overhaul it with extra params for shaping the noise mix response? i think.. "it would be worthwhile to do as a technical instrument, even if uninformedly, eg. scale, distort, limit the zero crossing ratio count mapped to each frame.." but limit accessibility and fun of this thing. plus a lot of the horrible work for adding modulation dests.

loads of stuff one can do.. separate formant shift for voiced and unvoiced.. useless.. happens with the expand/compress thing.. would be a nice basis for a deluxer synth with multiple oscillators and filters. engine is liquidy realtime.
you come and go, you come and go. amitabha neither a follower nor a leader be tagore "where roads are made i lose my way" where there is certainty, consideration is absent.

Post

had a go at making teh noise less repeaty,

i've done a bit with "rendered noise to looping wavs" in teh passed and thus accepted a bit of "repeaty" as intrinsic to teh method.. but those 16384 tables are def. useless if you wanted to eg. pitch a voice as breathy and stretch it out real slow.

i tried crossfading between two tables, awful, tried larger tables, and was about to try a granular table read method when i noticed that 65536 tables were rendering 4x smoother. so best solution for cpu in use, no crossfading, one table of 65536 samples. alright buy me.

except, the analysis/table writing part is now 4x longer, so it takes about 4-5 seconds to analyse a 2 second wav... so the currently available build is much more fun to use. i don't think i can squidge download links two editions on one product page, so due to tremendous interest, i'll keep it for myself. if some geriatrics get on about it, maybe i'll u/l it as "long lifter" or such.
you come and go, you come and go. amitabha neither a follower nor a leader be tagore "where roads are made i lose my way" where there is certainty, consideration is absent.

Post Reply

Return to “xoxos”