| Author
|
that's impossible
|
ND
ProSect
Started Topics :
44
Posts :
919
Posted : May 23, 2004 01:28
|
|
zooter
IsraTrance Junior Member
Started Topics :
44
Posts :
771
Posted : May 23, 2004 05:00
|
|
some kind of a hoax? |
|
|
Amygdala
Amygdala
Started Topics :
12
Posts :
175
Posted : May 23, 2004 10:19
|
I believe it could work to some extent. The splitter is based on Fast Fourier transformations, that "detects" the partial frequencies of a sound. This combined with some pattern recognition and simple arithmetic could possibly make the splitter work. Since sounds are usually built up from a fundamental frequencies and overtones, these can be detected and split up.
I'm dead sure that it won't work in every case though, but that could create some interesting effects, if you for instance try to split noise 
- Amygdala |
|
|
|
medir
Inactive User
Started Topics :
113
Posts :
1193
Posted : May 23, 2004 11:28
|
...yes, theoreticaly it s possible...n a guy from the mit isn t bad i guess...
however...where can i get a demo ?
bomski
 experiment !
make it your motto day and night.
experiment,
and it will lead you to the light.
the apple on the top of the tree
is never too high to achieve,
so take an example from eve...
experiment ! |
|
|
|
motorik
IsraTrance Junior Member
Started Topics :
14
Posts :
153
Posted : May 23, 2004 11:34
|
Yes it won't work in every case.
But it's possible....
Good sound engineers can do it manual to parts like vocal and the really good ones can do it even to other musical parts.
Read the exact way to do it once...i don't remember the whole method but (for example vocal) it's more or less made by separating R and L to mono then taking the mono track and puting band eq to get the vocal to cut all the rest after that you take the same sample invert it (flip the phase of sample) and than do band reject for the vocal frequency...than you export the two samples together.
thats should create anti phase and leave almost only the vocal to be heard...
the noise that left can be cleaned with gates.
This plug in might do all this process within itself....so it's possible.
I Read it long time ago and i'm not sure i gave all the info....i will try to find a link for this.
Salute
motorik
"one good thing about music is when it hits you , you feel no pain" - Bob Marley |
|
|
|
Kaz
IsraTrance Full Member
Started Topics :
90
Posts :
2268
Posted : May 23, 2004 11:41
|
It can be done - just that any compressed loop will be at least partially killed by this action (a compressor is a destructive process), reverbs will make it more confusing, and I seriously doubt this program can handle complex processes or mixing 'mistakes' properly. Or analog sound for that matter.
There HAVE been a few attempts at this before, and they have all failed miserably.
http://www.myspace.com/Hooloovoo222 |
|
|
|
ZilDoggo
Started Topics :
4
Posts :
663
Posted : May 23, 2004 21:57
|
i think it must be EXTREMELY hard to do it right.,
i'm not talking about the karaoke way (killing everything but the middle) which is a cheap trick.,
i'm talking about taking a sound out of it's context (the mix).,
the problem is that the program will need to understand human hearing (spacial impression, frequency, timing) .,
then it will need to decide on a saple to sample basis what frequency belong to what group of frequencies.,
i think that the complexity of the mix will determine the results.,
a mix with 6 instruments that play different tunes is not so difficult but what about an orchestra where a lot of instruments play similar tunes that are harminically related and are in a big big reverb.,
that's a different thing.,
and how about music that is no so harmonically related?.,
so there are a lot of situations where any kind of algorithm will start failing.,
i can tell you that to be able to take a sound out of it's context is something almost magical., 
and yeah, where IS the demo?.,
greets.,
aka., |
|
|
|
blueOrb
IsraTrance Full Member
Started Topics :
100
Posts :
1698
Posted : May 24, 2004 08:02
|
|
Amygdala
Amygdala
Started Topics :
12
Posts :
175
Posted : May 24, 2004 09:20
|
Hoax or not - for special cases it can be done. The way it's presented on the website above (hoax or not) it's based on FFT, which means it's not based on sample-by-sample.
Rather it's based on frequency-by-frequency. After the FFT (Fast Fourier transformation) you have information about particular frequencies, and can calculate which underlying sinusoids contribute to the entire mix - their exact frequency can be pretty accurately determined, and information about amplitude and phase is implicit in the representation of the sound - so spatial information is consistent. The tricky part is to determine which frequencies are fundamentals, and which harmonics belong to which fundamentals - but it can be done in simple cases.
Whether the SoundSplitter presented above exists or not, I know for sure that it's possible - I've written a program myself that works perfectly, if you are satisfied with individual sinusoids, and don't care about notes' overtones (pretty dull, but it works)
- Amygdala |
|
|
|
br0d
IsraTrance Junior Member
Started Topics :
12
Posts :
355
Posted : May 26, 2004 10:13
|
Reverb, ambience, harmonics, diffusion, etc will always prevent this sort of effort from ever being totally effective, obviously.
They might be better served to apply this sort of FFT technology to an automatic EQ that bandpasses around the fundamental frequency of an instrument so people can rapidly shape tracks without having to torture their ears by jacking up the gain and Q to sweep and locate a fundamental. It could also obsolete frequency charts for those of us anal enough to use them.
Automatic track EQ would be badass workflow tool to have.
It might also help if the interface didn't look like Soundforge 4 |
|
|
|
Amygdala
Amygdala
Started Topics :
12
Posts :
175
Posted : May 26, 2004 10:56
|
Why only accent the fundamental...? I think that would make sounds kind of boring and uniform. What makes a sound have a certain timbre, is anything but the fundamental. Some melodic sounds don't even have a fundamental, even though it's perceived...
But the idea can work for some applications. I think I'll build one, when I'm done with my thesis
-Amygdala |
|
|
|
mfa
Started Topics :
8
Posts :
168
Posted : May 26, 2004 19:32
|
|
so, where's a link to download ? |
|
|
|
br0d
IsraTrance Junior Member
Started Topics :
12
Posts :
355
Posted : May 27, 2004 08:17
|
Quote:
|
On 2004-05-26 10:56, Amygdala wrote:
Why only accent the fundamental...? I think that would make sounds kind of boring and uniform. What makes a sound have a certain timbre, is anything but the fundamental.
-Amygdala
|
|
Because the Q is the real adjustment, the fundamental is only the starting point. Many times sounds are HPF'd and LPF'd to fit them into a busy mix, and having the fundamental already centered for you would be pretty convenient. Often times mix components sound pretty boring when soloed. Ideally either the HPF or LPF could be bypassed. |
|
|
|
Kaz
IsraTrance Full Member
Started Topics :
90
Posts :
2268
Posted : May 27, 2004 10:47
|
I'd be very impressed by this tool if it had 100 times more options to configure by setting marker points on the wave of the original file, and a few artifact removing/creating options in order to keep the db limit. Of course, with psytrance productions where the bassdrum is often left at -0.1db because everything is ubercompressed, the only sequence you might sample properly with something else playing is a bassdrum.
Even then, it would only give a rough outline or a few 'good guess' options, but even those would be more than enough - and even then I'd know that the sound would probably fit only in the 'slot' allocated to it in the original track. http://www.myspace.com/Hooloovoo222 |
|
|
|
