this is a patch I've been working on for a while now, I've uploaded progress a few times. This recent version has a save parameter feature, and the controller permutations have been tweaked.

you can automate any of the parameters by sending messages like "SampleHold $1" or "UpSample $1" to the 2nd inlet. I swear I'll get around to uploading some music soon, so you can hear where I'm trying to go with this.
UPDATE: second link is the new and improved.

http://www.pdpatchrepo.info/hurleur/specmash.pd

Hi sunji, "countto" is missing.. could you please post it (or where I can find it)?

Thx, AlbertoZ

sorry, missed that. Here tis.

http://www.pdpatchrepo.info/hurleur/countto.pd

54 downloads and no one's tried it out? back to the drawing board...

Sunji, I shall have a look at your patch with excitement, tomorrow.

i played with this ages ago on one of the earlier releases, and just got noise. but now i had another go, and found some really nice tweaks.

with the samplehold really low*, then modulating the window size between halfway and full has a good effect.

*(samplehold needs to be low, but but not totally down, because then it is zero.)

with the window slider at about 70%, modulating the samplehold slider has a really good effect too. gets really bubbly with overlap added.

i made 3 modifications.

(1) changed the samplehold slider's output to log, between 50 and 44100.

(2) added [t b f] objects after both the [r $0-Overlap] and [r $0-Upsample] objects in the process subpatch. otherwise the overlap and upsample sliders don't do anything until the window slider is moved.

(3) added an inlet to reset the phasor~ phase. the shifting phase of the phasor gives really nice subtle changes to frequency output, but for times when I need sounds to be tight, resetting the phase is important.

anyway, really cool simple process, and i definitely hope to use this for making some samples and trax.

Sunji this is fantabulous, I really need to get into the world of FFT.

I have exactly the same thoughts as Hardoff.

I was wondering if there would be a way to have those amazing sounds without all the distortion. Then I noticed I was way overdriving the signal. But that made it all the more interesting, since turning down the gain to acceptable levels created only tremolo like sounds. Tried simply compressing it to bring out those low signal sounds (since that's what I thought I was enjoying from the overdriven signal (the fact that it compresses everything)) but wasn't really my ticket.

Is there a way to bring out those in between sounds (maybe overtones?) without suffering the world of square waves. tanh~ isn't really doing the job

simply feeding an [osc~] is great!

This is just a repost taking hardoff's suggestions, and also reduced it's screen footprint.

Also like to point out it's automatable- 2nd inlet accepts values like "Gain $1" and "Window $1". That's something I do in all my patches- a slider can always be called by it's name for automation. I'm thinking about unifying all the inputs to ranges of 0-127, for ease of interacting with midi. Any body have any reason not to?

http://www.pdpatchrepo.info/hurleur/specmash.pd

even better!

Try sucking inwards infront of a mic. You get the most amazing lovely, modulated high-end weird-ness with none of the overpowering full band noise that blowing or [noise] brings.

Then go between noises like sssssss(suck) and boooooom.

If the latency could only be improved and the transition between clean and fucked up, made smoother, this would be a great live tool

when you use large windows you necessarily introduce latency, to the tune of double the frame sample size.

If I could reselect the sample rate in the patch, then I could sync the latency with a rhythm. But there does not seem to be a way to alter the samplerate in a patch subwindow, like you can affect block size with [switch~] and [block~]. anyone got a hack for this?

there is a patch in the browser on upsampling....

browser > 3.audio examples > J07.Oversampling

but that seems to use block to do its work... is that of any help?

omg

this thing is amazing!

I didn't realize the full potential until I started cutting down the volume of the sound going in, at really low levels, it does amazing things.

nice work

When you get something you like, post it and let me know!

hey many respect for this patch! I really like it!

So the base of this whole thing is the samplehold object?
I understand what the object itself does, but what that actually means for the rifft object is not clear to me. It delays simply the data of the rfft right? So why is the sound altering in the way it does? During the last days i tried to get into FFT and i understand the basics of it i would say but this surprises me.

I did some control additions:
For example i divided or multiplicated the base samplehold value by 2 every time a press a button.
Also adding or subtrating values by ten is interesting: it lets the structure rise or fall respectivly.
And i added a counter which counts up or down on the base of the samplehold value.

I found it useful for slightly altering the rhythmic and tonal structure...

Is there any way to get rid of the distortion on some tones??
Anyway great work!

Did you also come up with other things using fft objects? Would be very interested!

Thanks for your feedback Rancine!

the distortion is kind of integral to the sound. The [rfft~] [samphold~] [rifft~] chain is sonogram smearing. I mean it distorts the image of the sound spectrum.

I will put thought into other avenues of fft sound manipulation. This one seemed obvious, but there are other great sounds waiting out there no doubt.

so here's a song a friend recorded, and I spectralmashed. Just a por ejemplo.

http://www.pdpatchrepo.info/hurleur/satansarms.mp3