Forums > MaxMSP

OT- It's impossible, but he's doing it.

March 11, 2008 | 11:18 pm


March 12, 2008 | 1:25 am

WOW!

Is that in the current version? Does it actually work as well as it demos on video, because we all know video is reality. :)


March 12, 2008 | 4:05 am

well, the luddite in me is incredibly pissed off (changing the key of my samples?)
but the technocrat in me is f’n blown away.

anybody have a clue how this thing is possible?

oh, and what’s meaty?


March 12, 2008 | 4:41 am

I would like to see/hear this with other instruments or a chorus…,
maybe with guitars it is easier, because they have a nice overtone
spectrum.

first you need a good pitch shifting algorithm. you don’t have to mess
around with phases because you grab the note as a hole, and maybe don’t
need to split it into separate fft-windows. I think one part of a good
pitch shifter is the ability to separate the basefrequency from the
formants to keep the sound quality of the note. but have not done this,
so don’t know how difficult it really is.

the really difficult part is the separation of a chord, where you have
to look at base frequencies and find out which overtones belong to which
groundfrequency. with strings and maybe also with other instruments you
can predict the overtones and use this information to grab the correct
frequencies for shifting…

just a guess.

marius.

Robert Ramirez wrote:
> well, the luddite in me is incredibly pissed off (changing the key of my samples?)
> but the technocrat in me is f’n blown away.
>
> anybody have a clue how this thing is possible?
>
> oh, and what’s meaty?
>


March 12, 2008 | 4:58 am

Quote: marius.schebella wrote on Tue, 11 March 2008 21:41
—————————————————-
> the really difficult part is the separation of a chord, where you have
> to look at base frequencies and find out which overtones belong to which
> groundfrequency. with strings and maybe also with other instruments you
> can predict the overtones and use this information to grab the correct
> frequencies for shifting…

Yeah, I’m betting this works really well for harmonic sounds but I guarantee there are musical signals where it will fall apart. Like many-layered noisy pitched percussion with inharmonic spectrum.

I like that quote "I began to see that what doesn’t work in theory can still work in reality." A lot of times people think because a problem is really, really hard or because it is "theoretically impossible" that no solution can be found. There are always heuristics and approximations! In the past I worked at a company that made a good deal of money selling software that solved a problem that is theoretically proven to be impossible to solve with a computer. In the context of software, theoretically impossible (i.e. incomputable) just means you cannot solve the problem with 100% accuracy. It doesn’t mean you can’t solve a problem with 95% or even 99.99% accuracy if you take lots of educated guesses…


March 12, 2008 | 9:59 am

Must say i was pretty impressed, and i’m not even into audio editing. There must be situations where this doesn’t work perfectly, such as me adding really soft overtones..but the examples were pretty convincing.

And i have my doubts about the next phase of their plan…


March 12, 2008 | 3:18 pm

having worked with celemony and autotune a bit, i can say that this i
infuckingcredible. now can it take a chorus singing the same note and
separate it into separate voices?
b

On Mar 12, 2008, at 5:59 AM, Bas van der Graaff wrote:

>
> Must say i was pretty impressed, and i’m not even into audio
> editing. There must be situations where this doesn’t work
> perfectly, such as me adding really soft overtones..but the
> examples were pretty convincing.
>
> And i have my doubts about the next phase of their plan…
> –
> SmadSteck – http://www.smadsteck.nl
> Hard- and software for interactive audiovisual sampling

bruce tovsky
http://www.skeletonhome.com

"Reality is that which, when you stop believing in it, doesn’t go away."
Philip K. Dick


March 14, 2008 | 1:02 am

This is simply incredible. Im freaking out with this program!!


March 14, 2008 | 1:40 am

it is really amazing… I was taught this was physically impossible, and PK has just gone and made it work.

In this video he does a string quartet… force Mozart minor in an instant:

http://www.sonicstate.com/news/shownews.cfm?newsid=6281

But I do wonder about very rich things, like a string section… it might be a bit more difficult… nonetheless, it’ll be so much fun to play with when it’s released.

Can’t wait.


March 14, 2008 | 2:03 am

I’ve searched google for more about it, and found a video interview (soundonsound.com) with the guy who invented it. It seems that one of the problems of this technology is that can’t divide instruments playing the same note.

Probably because it makes some kind of analysis based on tonic-overtones to extract the sound, so it’s pure logic: same note, same overtones (more or less).

I might have listened bad (view the video and correct me if so) but the guy said that this plugin also eliminates noise from the audio. It makes sense, because it doesn’t belong to music overtones, so it could be filtered out, or recognized as another source and deleted.

Great program by the way, shame it won’t sale until end of the year :( (at least)


March 14, 2008 | 10:00 am


March 14, 2008 | 1:33 pm

I was never able to get the polyphonic pitch tracker
to work. Maybe I should try it again.


March 16, 2008 | 8:20 am

Dan Powell schrieb:
> I was taught this was physically impossible,

But your ear would have proven that statement wrong already, wouldn’t it?

Jorge schrieb:
> It seems that one of the problems of this technology is that can’t
> divide instruments playing the same note.

That’s spared for the next "he did it again"… ;-)

> Great program by the way, shame it won’t sale until end of the year
> :( (at least)

That will give us the chance to create a Max external before he has
streamlined it for the big market… ;-)

Stefan


Stefan Tiedje————x——-
–_____———–|————–
–(_|_ —-|—–|—–()——-
– _|_)—-|—–()————–
———-()——–www.ccmix.com


Viewing 13 posts - 1 through 13 (of 13 total)