Extracting guitar and keyboard from the record

[Mt Umut Sarac]

I want to create a samples library of Steve Vai and Emerson Lake and Palmer.
And I want to create a Oberlinger Organ Sampler.
I need to eliminate drum , vocal and bass sounds from the original track and work with the extracted guitar, synthesizer , organ left or right hand record file.
Which computer technique , software or code do this ?
Which conference publish techniques about this ?

[tedluk]

I don't think there is any possible way to do this. Certainly not to a degree where you could use the resultant audio for samples.

Unless the audio you want to use is panned separately from the other instruments, it's going to be impossible to isolate the keyboard and guitar from the other instruments completely. If the parts reside in their own frequency range you could try using some radical filtering to eliminate the other instruments, but I doubt that the result would be musically pleasing. Even then you're going to get some bleed through of the other instruments.

Maybe someone else knows a way to do it that I'm not aware of.

Of course if you can get your hands on the original multitrack recordings then it would be a lot simpler.

Ted

[Farview]

Just call Steve Vai and Greg Lake. I'm sure they would be happy to give you the master tapes so you can profit from their hard work. (and the producers, engineers, etc.)

Now, if you found the instruments, you could just get some mics and and stuff together and sample the heck out of it.

[stonepiano]

Originally Posted by Farview

Just call Steve Vai and Greg Lake. I'm sure they would be happy to give you the master tapes so you can profit from their hard work. (and the producers, engineers, etc.)

[Mt Umut Sarac]

Originally Posted by tedluk

I don't think there is any possible way to do this. Certainly not to a degree where you could use the resultant audio for samples.

Unless the audio you want to use is panned separately from the other instruments, it's going to be impossible to isolate the keyboard and guitar from the other instruments completely. If the parts reside in their own frequency range you could try using some radical filtering to eliminate the other instruments, but I doubt that the result would be musically pleasing. Even then you're going to get some bleed through of the other instruments.

Maybe someone else knows a way to do it that I'm not aware of.

Of course if you can get your hands on the original multitrack recordings then it would be a lot simpler.

Ted

I found a software from blazeaudio and there is a discussion board also.
They say soundforge and eqs help to do this work.
How can I do this radical filtering with the own freq range ?
Which software do this ?

[Farview]

You can't. It won't work. Stop wasting your time thinking about it.
BTW Steve Vai's guitar sound is not that hard to do. The old organ sounds are going to be a pain, because you will have a hard time finding the organ.

[gordone]

I wrote some Java code to do this.

Music song = new Music("myfile.wav");

Music keyboard = song.getKeyboard();
Music bassDrum = song.getBassDrum();
Music cowBell = song.getRequiredCowbell();

keyboard.writeToFile("keyboard.wav");
//and so on..

You would just have to implement the get*() methods in the Music class, and then you'll be all set!

[tedluk]

Man! You gotta love that Java!!



Ted

[Mt Umut Sarac]

I read a phd thesis of bernd schoner from mit.
Name is digital stradivari.
You extract pitch , loudness , brightness and spectrum peaks from the record and connect them with artificial intelligence.
Spectrum peaks made me to think about them. If you handle the peaks of spectrum , loudness , brightness and pitch graphics , you can see when a new sound come in to the mix and get out from the mix.
If we say , whenever a new sound come in to the mix , the spectrum peaks - some of them - move at the same time , we can extract the newcomer from the peak graphics page.
What do you say that ?

[Farview]

If you know how to do it, why are you asking us? That sounds like the kind of thing the FBI uses to clean up audio tapes and such. I don't think it will give you what you are looking for.

[tedluk]

Originally Posted by Mt Umut Sarac

I read a phd thesis of bernd schoner from mit.
Name is digital stradivari.
You extract pitch , loudness , brightness and spectrum peaks from the record and connect them with artificial intelligence.
Spectrum peaks made me to think about them. If you handle the peaks of spectrum , loudness , brightness and pitch graphics , you can see when a new sound come in to the mix and get out from the mix.
If we say , whenever a new sound come in to the mix , the spectrum peaks - some of them - move at the same time , we can extract the newcomer from the peak graphics page.
What do you say that ?

Good luck! Check back with us in a couple of years when you get it all worked out. In the meantime, the rest of us will be making music.

Seriously, WHY are you wnating to waste all this time screwing around with such a hairbrained shceme? You want organ samples? There are thousands of them out there. You want guitar? Same thing. On top of that, there are all kinds of different synthesis methods that can recreate guitar and organ such as physical modeling.

We've all been having a little fun at your expense in this thread because what you want to do is, at best, going to be enormously difficult and almost certainly not result in anything usable. You'd be much better off learning how Steve Vai creates the sound you like and learning to play better. Same holds with ELP!

Ted

Ted

[bennychico11]

Originally Posted by Mt Umut Sarac

I read a phd thesis of bernd schoner from mit.
Name is digital stradivari.
You extract pitch , loudness , brightness and spectrum peaks from the record and connect them with artificial intelligence.
Spectrum peaks made me to think about them. If you handle the peaks of spectrum , loudness , brightness and pitch graphics , you can see when a new sound come in to the mix and get out from the mix.
If we say , whenever a new sound come in to the mix , the spectrum peaks - some of them - move at the same time , we can extract the newcomer from the peak graphics page.
What do you say that ?

you have it completely wrong. you need to re-read the thesis. First off, the correct name to his patent is the Digital Stradivarius. And this has notthing to do with extracting individual sounds from stereo recordings. His ideas was micing an instrument and recording both it and the movements of the player. Putting this data into a computer, he was then able to recreate the sound of the instrument without the instrument. In other words, someone who didn't have a Stradivarius violin could use a computer to help replicate the exact sound of one.
http://xenia.media.mit.edu/~schoner/press/FRAMES99.html
(original site: http://xenia.media.mit.edu/~schoner/ )

straight off his website at MIT

Now back to what you're wanting to do. You can't do it...at least not perfect enough that you can't hear a single other instrument in the background or ruin harmonics of the instrument. First...try recording a piano player, and then record a guitarist separate. Look at both of them through a spectrum analyzer. Notice how frequencies are recorded all the way from 20hz to 20KHz and beyond?? And in both instruments?? Even if the instrument didn't play that note, it's still there. That's the principles of sound.

Of course, there are devices or programs you can get that will try to extract a certain instrument, but i've never seen one that can do it without flaws. You're going to lose some of the important frequencies. Good luck.

[Farview]

You really would have a better chance of getting Steve Vai and ELP to show up at your house for you to sample them. The only equipment you need for that is your checkbook. No need to rely on anyone from MIT.

[tedluk]

Originally Posted by Farview

You really would have a better chance of getting Steve Vai and ELP to show up at your house for you to sample them. The only equipment you need for that is your checkbook. No need to rely on anyone from MIT.

Hey! I'd pay to see that show!


Ted

[Necrothing]

Theres a guy from Turkey (I think) who plays just like Steve Vai, and he just uses Cubase VST with a couple of plug ins (1 being a JCM 900 emu) and he manages to replicate the sound of Sir Vai himself. You might talk him into doing a sample for you, he's on a forum called KVR or something.

Failing that, buy a JCM 900, an Ibanez Universe and an Eventide H3000. Then spend the next 25 years learning how to play.

[Farview]

Steve has his own signature Carvin amp.