sitar model (download available)

[xoxos]

a few months ago i posted these files of a sitar model:

http://www.xoxos.net/temp/sitar.wav
http://www.xoxos.net/temp/sitar2.wav


i've been promising it for a few months and occasionally people have asked for it, so here's the official word on it:

no sitar model vst

the distinctive 'twang' is produced by interaction between the string and the curved bridge. when the string vibrates, it hits the bridge and shortens the string length.. if you listen to it "informedly," the sitar tone is ~a musical tuning of a 'vibrating against something' sound..

computationally, the fact that the string amplitude affects the string length in realtime is problematic, as the current string length is required to compute the output/amplitude..

you may remember another audio file i posted, wherein the string would 'catch' and sound infinitely. given my development method, it was through trial and error that i found coefficients that produced, under circumstances heard in these two 'good' demos, emulative tones, ie. twangy, with a decaying tone.

ftr, in these examples the string is oversampled twice, and the output is used to compute the string length for the next sample..

basically, it was a stroke of luck.. increasing the oversampling had the same effect as lessening it.. a less emulative timbre, more plucky than twang.. aamof, anything i did worsened the model.. emphatically, i found that the plucking impulse only produced the amount of twang heard within narrow guidelines, so that changing the impulse did not alter the timbre as predictably as it does in other karplus-strong models. adding high frequency (eg. noise) would diminish the jawari effect but not add any brightness to the timbre.

so that sucked.

using eq after the fact did help but around 4-5% cpu per note with no flexibility on impulse shaping beyond amplitude.. the model did not offer any timbral flexibility over sampling (eg. amplitude only) so there's no point in me spending a week to accoutre the vst.. if you're disappointed, use sitar samples and lowpass around 4k.. that sounds about the same as what it would have produced, at best.

i'll probably put together a resonator/sympathetic string effect using a few string models.. i tried using a 2nd instance of the string as such and it sounded alright

development -> unpredictable outcome..

[stanlea]

alright in case you couldn't make a sitar, just make a xitar

[xoxos]

why?

i'd probably use it once in my history of recording, if that, maybe never, and i've got a load of things i could spend that week on that might actually do people some good. just because i make commercial vst it doesn't mean that people buy them.

it's like i have this wild ambition to be a person that codes things with FFT but code things that i would code instead of what other people would code.

call me crazy :p

you know i've probably spent a year of my life already making up for not taking calculus in school.. trawling through the web trying to figure out the things i'm supposed to know in order to understand this or that. time is fortunately one resource i have, so i don't like to squander it

[xoxos]

http://www.xoxos.net/temp/sitarwres.mp3

as you can hear, while it is evocative of a sitar, it just isn't sitary enough, enough of the time and sounds more kotoey or something.

there are three commuted resonators here, which runs about 5% of my 1.6g cpu. as said, the main put off for me is the lack of the ability to infuse high frequencies and flexibility in the impulse. the jawari as i've implemented it translates it into less twang.

please excuse my playing, i have to use the visual keyboard in synthedit/qwerty keys.

[Jafo]

re: the latest demo:

I especially like how the jawari and resonance kinda fade in. You have a very responsive model indeed. However, I think you might want to let them take longer to fade in; or maybe it's that their attack seems to be tied to the velocity in an inverse manner -- softer notes seem to fade in more quickly. But then, I'm probably not listening closely enough.

Also, it seems like the buzzing is in too low a register, or is pitched an octave too low. It's a nice bassy buzz, but the real thing or samples thereof buzzes at higher pitches.

Overall, the sound is fuzzier, deeper, and more square wavy than I expected.

Still, this is very nicely responsive. I've got samples and String Studio patches that get the sound, but yours pretty much kicks their collective ass as far as feel goes. I'd like to urge you to keep at it, because I think you'll have something truly astounding when you get the sound right. Gotta be hard work, since you're clearly trying to model how the tone and harmonics change throughout the note's lifetime, rather than just taking one picture and stretching it out.

[xoxos]

my surmisal is that the 1 sample latency in the jawari (explained above) is why it needs a certain range of low freqs in the impulse and why higher freqs in the impulse reduce the jawari effect.

(the output could be lowpassed for processing the jawari response but then there's a phase issue and if you're going to 'part' it one might as well use additive synthesis instead of procedural synthesis).

jafo - yeah.. there's nothing quite so antithetical as a sitar swell sample imo. it's supposed to sound trancendental and alive, but the 2nd time you hear the exact same swell (et alone the 3rd itme..) it gives me the creeps. unfortunately, being such a resonant form with laods of high freqs bouncing about, i think my attempt to model a low cpu solution was somewhat optimistic.

[Shabdahbriah]

... i'll probably put together a resonator/sympathetic string effect using a few string models.. i tried using a 2nd instance of the string as such and it sounded alright

I will BUY THIS immediately!!!

I am working on a ton of 'modeled' and twisted "string" sample banks in/for the sfz format.

(assuming you relaease it at or shortly after the 1st of any given month though)

[xoxos]

you know your way around synthedit some, right?

[Shabdahbriah]

you know your way around synthedit some, right?

I have had "some" lucid moments, yes... but TBTH I got considerably thrown off track by the known and potential compatibility issues with the major SE "update" a while back, and all of the tools (modules/prefabs) I had purchased and been accumulating, which suddenly were becoming obsolete.

Though I have since gotten most of the full-package 'updates' (CK, DH, EVM, etc), I've yet to load any of them, as I've busied myself with other projects and "learning-curves" while I continued with my 'foundational' waveform and sample-based 'instrument' mutations, mixed with innumerable and persistent family crisis's, and the occasional quixotic world domination scheme.

I still want to pursue converting my samples (for SE) for use in some guys "text" module BTW

[xoxos]

aah - i'm still using a pre-1.0 release.

with all respect due to jeff, i haven't found the documentation yet (i have the wazoo book, which, again with apologies, i found incomplete and unstructured) to appreciate the subcontrols. almost everything i've built with subcontrols, including other peoples' resources, is unresponsive (eg. takes several tries to select a control) so i stuck with what works for me, ~.96 something. vst sdk is next.

i haven't had any incompatability reports due to my .sems, probably because i don't try and do anything unusual with memory allocation, dependencies, et c. i cna send you the module and the .96 release if you like, easier than mucking with it myself :p

[Shabdahbriah]

aah - i'm still using a pre-1.0 release... i cna send you the module and the .96 release if you like, easier than mucking with it myself :p

That would be AWESOME, thank you.

I have SE 1.0170 loaded currently, and I haven't had any "compatibility issues" thus far with the very limited number of things I've tested, and I'm pretty sure I have several pre-v.1 SE's on disc somewhere, if not already installed on one of my other machines.

Please PM me a link if you would, and a quickie 'screen-grab' of any kind of (even extremely sophomoric) project showing its to/from connections, would be greatly appreciated. My synaptic/memory issues which pertain to 'connecting-the-dots' have rendered me about 90% "visual"[ly] dependant.

I'll gladly forward you any 'samples' or SE experiments I make using it.

Re: Converting .wav to text, for use in an SE 'module' for samples/Osc's etc., I recall you either made or were toying with the idea of making:

Would this work?

Watex

('bout half-way down the page)

Thanks again

[xoxos]

i use wavosaur :p

edit..

http://www.xoxos.net/temp/sitarwres.mp3

opened old project and decided maybe it wasn't so bad.. not great or anything but audible.

soooooooo... i'm going to take a day and gui it up super easily and release it

iow, brb

[xoxos]

http://www.xoxos.net/vst/mp3/plingtar.mp3 (same mp3 as post above)
http://www.xoxos.net/vst/plingtar.zip

don't expect it to play nicely. i took the in-process file i had and split it into an instrument and a resonator effect. neither are patched, i plugged one into the other to make sure it worked and the effect could be mixed way down for starters... otherwise it's the same settings as in the mp3 :/

if you're looking for plucked string weirdness, try the 'deva' plugin i released earlier this year.. a karplus-strong model with a few mass-springs along the length that modulated length.. i don't have the frequency > gain responsivity quite smoothed out so it was a bit explodey but often sounded the closest i've heard software get to kicking a spring reverb fender amp splat (i don't know if this is still online..)

anyway, sorry it doesn't sound like a sitar

[xoxos]

and hope you like the "would" look

[Musical Gym]

thanks for releasing. I like the look.