fft and granular, how do you guys use it?

[kensuguro]

Here's something music related.. FFT and granular... very different tech, but to in practice, they seem to be used in similar ways. A lot of demos I see just pull stuff very long. Hey, you can stretch speech super long and even "freeze" it. Both of them do it, and they sound sort of similar. Strange metalic crazy sound that I guess you could filter, pass through some reverb, mash it up and call it some sort of ambient drone. But that gets old very quickly. At least for me it did. I mean, it's like a blender or vita-prep.. Everything gets blended into a puree and it's all just this undiscernable mush.

I guess the more practical application is beat slicing, and both are used for beat slicing. That's practical, and useful. So I'm not concerned with that. I'm thinking more in terms of synths.

Granular seems to be used in pad sounds more so that other type of sounds. Not too sure about fft. I have Alchemy, which does both, but I'm really not sure which is being used for what presets. (I can look, but it's not obvious from just listening) All in all though, to me, granular seems to have found its place with pads.. glass shimmers, metal tingles, metallic hums are iconic.

So what about fft? Great for formant shifting and time domain manipulations.. Is it just a fancy sampler? Sure, you can sample a 1 voice and fft the hell out of it to get range. But we don't need to do that. You just go buy a 500gb library. As a sound generation method, I personally haven't found a creative use for it yet. I guess you can blend resynthesis and kind of morph between samples like kyma does.. but that's also gimmicky.. I mean, how many times can you morph a man's voice into a lady's voice.. or into a violin sound.. That was cool 20 years ago when phasevocoding became known. It's interesting conceptually, but musically, it's really only about as cool as a crash cymbal. Autotune also uses stft, but again, that's different from using it as part of a synth signal flow.

Is fft better suited as a sound design tool? Maybe it's not meant to be a synth thing. What do you think? How do you use fft in synthesis? How do you use granular?

[valis]

When I was playing with granular in Matlab & CSound I found STFT is actually more fun for me than most other fourier transforms. as you can tune the window size (and windowing function) to have a rhythmic 'warbling' that would be obnoxious in timestretching normally, and get extremely long stretch times. I think there might be some LADSPA plugins that do this. Later there were also techniques to vary the time window size but I never played with these. It's worth noting that STFT enables 'phase vocoding', which I believe is used both in NI's Traktor for timestretching mp3's and in Auto-tune.

Lately I've been playing with a few Reaktor ensembles and the following Mac apps:

http://xenakios.wordpress.com/2012/10/1 ... -released/

http://argotlunar.info/

[kensuguro]

ya, i think most audio applications are stft since you need to window blocks with FFT, in which case that's really stft.

[Mr Arkadin]

Not sure what you guys are talking about, fft stft? What's all that even mean?

Anyway I've used a bit of Granite from time to time, here it is in a remix I did (particularly the end sequence):

http://soundcloud.com/luke-antony/scann ... l-endure-1

[hubird]

I forgot about the granular when listening, except the very intro.
A top mix.
Not sure about your part in re-mixing, but it's a perfect mix for the music.
Hats off.
Damn, must check the granularising again

[Mr Arkadin]

A top mix.

I dunno. I always felt it needed more feedback technic Too much Marshall bullshit technic in this mix.

Not sure about your part in re-mixing, but it's a perfect mix for the music.

Not quite sure what you mean here? Anyway, the only thing that was used from the original Gary Numan track was his vocal (also granulised on the chorus) and a bit of granulised-beyond-recognition piano at the end. Everything else is Scope instruments, BFD2, guitar and bass plus the Granite VSTi.

[hubird]

you know what?
I skipped the same 'joke' (feedback) , in order to keep the compliments clean

I thought you called it a remix, ...edit read your upload, ok then.
Didn't recognize anything tho, I didn't follow Gary Newman
Really great mix then, I don't see myself doing this.
The base sounds wild, and life. Fits perfect.

[Mr Arkadin]

The remix was from this track:

http://www.youtube.com/watch?v=NYv2VrEhmBY

These were the stems I had:

http://soundcloud.com/bttp/sets/gary-nu ... nner-stems

The base sounds wild, and life. Fits perfect.

Oh, I forgot the bass synth that appears just after the intro is my Taurus 3 pedals.

Really great mix then, I don't see myself doing this.

That's because you do not know the true path of feedback technic.

[hubird]

1. If I only knew what's he doing with that feedback...
2. wow...will check now
3. I stand corrected. Never knew he made this. He knows to create his own world soundwise, it's great. . It's not my private taste, but he tries to avoid the obvious in melody lines and arrangements, even text, so it's great
Is the hard gating on his voice (in the or. track) intended, the attacks can be heard?

2006 if I'm right, so not that old.
It can stand the times easily.

I prefer your remix compared to the original, it's less style bound and more explosive

[jhulk]

granular is a lot like wavetable scanning synthesis

in that it takes a audio sample and turns it into grains then you can plot through these grains at what speed you like in any direction which gives ever evolving

sound scapes

now fft is a program method to change a wavesample into its basic form additive spectra and will give you single cycle spectra over time on the sample

then is played back as a additive spectra file this is better than doing it as a sample as if you transpose up a sample it gets faster and becomes munchcinised

but doing it additively it does it by spectrum so that its plays correctly over the whole range

harmor is another good example of an fft additive playback synth

it is also the method i use to create transwaves for ensoniq samplers

i use an fft program that slices the sample up into segments of additive spectra i then copy the additive data and make a wavesample of it then i piece each sample into a slot and additively interpolate to make a table

its also how i created the wavetables for the vectron and the korg delta modular synth for scope

[dante]

So a 'grain' is a section of sample that loops without a click - eg where the end matches up sonically with the beginning so perfectly that it loops at any speed without clicking ?

[jhulk]

yes a grain is like that

but it must do it like additive it must scan a sample and set grains to a fft additive wave as if it was done by samples it would would click unless each loop was
exact the the same size but with grains you can choose a few and it would still loop great with out clicks but it must use some type of interpolation algo todo this

bit like the uwave synths they create there waves in halfs in 8bits so you create the one have and the processor creates the other as a mirror image

which is great but in real world of additive is impossible to have identical side unless they are sin or cosin or a mixture as you add more harmonics like the pluse waves and others the + going ac wave is not the same as the negative but on the uwave they are a mirror image

and there wave tables are interpolated buy math and you can here it as the processor is to slow and so clicks and artifacts happen

on the scope module that uses the same system its made up of 32 single cycle waves and then interpolated between each one but because the scope dsp are fast you dont get the glitches or the aliasing like you do on the hardware they are smoother

when i listen to my waldorf wave or the uwave or my orange xt rack in the upper note the wavetables can be really harsh

but on the scope system they are smooth

the blofeld also uses the 32 waves interpolated system and largo

which is waldorfs version of komplexor which they designed when they went bust but on that they allowed user wavetables

like the blofeld

i actually sold my waldorf q because the komplexer was great and could read mq sysex and by loading a couple patches to the layers i could do all the sounds of the q but if they brought out the 16 filter version the first time round i doubt i would of sold it

[jksuperstar]

Grains are typically samples that are 20-100ms long, and have an envelope applied like an FFT window, with attack ramp, sustain, and release ramp. That means the beginning and end of a sample are 0 anyways. A granular sampler then overlaps or loops several grains, depending if you are speeding up a sample or stretching it out. That's why it can change the tempo without changing the pitch...the grains are played back at the same sample rate, just overlapped or looped to make the time difference.

An FFT based pitch shifter (ie sampler), uses analysis and resynthesis. The original sample is analyzed and transformed into the frequency domain, the frequencies are shifted in the spectrum, and then additive synthesis rebuilds the new output. That way you don't get Mickey Mouse effects, and can control the levels of each harmonic if you wanted to.

DSPs are perfect for either system, FFTs in particular, and it'd be nice to see more use of dsps to do these things in scope, to more differentiate it from native based synthesis/effects, which don't really have the horsepower or memory bandwidth to do the same tasks in realtime.