ASTROBOX.org - dedicated to H!P + J-pop fan remixes
 
 

Tutorial: Old method of extracting Vocals

CrappleJunkTown

Program used: Sonic Foundry Sound Forge 6.0
Intro
This is the harder, lower quality,non Utagoe Rip way of extracting vocals. Obviously for these reasons alone, I don't recommend it. This explanation is mainly for interest purposes only. It's based around the rule in sound physics that states that by taking the inverse of a wavelength and "mixing" it with the original, you create a cancellation, leaving the difference of the two. Therefore, theoretically, according to this, you should be be able to take a song and its instrument track, invert one, mix them, and magically an accapella track should be regurgitated, right? Well, under some circumstances, this is true. Unfortunately, due to slight differences in the mixes, this method doesn't always work. This is mainly due to slightly different tempos, which although possible to fix, is extremely tedious.
Steps
Okay - Firstly, you need 2 audio files, a song and its instrumental version. These need to be of fairly good quality to work well. Your best bet is to rip them into uncompressed waves if you have them on the source cds, but if not, you can try with high quality mp3s (at least 192kbps). Anyways, open them up in Sound Forge.

Secondly - You're gonna need to line up the files so they start at EXACTLY the same time. To do this, go to the options menu and under 'staus format', choose samples. Do this for both tracks. Next, zoom in on one track and find a destinct "spike" on the wave, where you need to click on the peak. On the second track, there should also be the same "spike" at the same part of the song. Click on the peak of that spike as well. Now you can even them out.

You should be able to see the sample number on the bottom of the track window for each song. You need to calculate the difference of these numbers. To even the tracks out, you have two options :

1- Add the difference to the track who had the lower sample number by going to the 'process' menu and choosing 'insert silence'. you then insert the difference number, and click for it to be at the beginning of the file. Click "ok".

OR

2- Select the track with the high sample number and go to the "edit" menu. Choose 'selection' and then 'set'. Under start, set "0". under end, set the difference number. Make sure the channel thing says "both". Click "ok"

Now, here's to see if this works. Select on of the songs by double clicking inside it's window. Copy it by going: edit -> copy. Next, go: edit -> paste special -> mix. Make sure the box next to the volume control under the destination title that says "invert" is checked. You can preview it if you want (it only gives a ~3 second sample), but otherwise click "okay" to see if it works.

Conclusion
This is not guaranteed to work with all songs, although I have had one superb result ("kokoro hitotsu" by Misia). Most of the time, it doesn't work very well at all, so I don't recommend it.