how to take the voice off?

[erzane]

I'd like to cancel the voice on a track, having only the music (I'm using Adobe Audition)
Thanks

[Cornucopia]

Barring special circumstances of cross-channel subtraction (ala destructive interference by changing polarities), or using a formant-freq. based specialty plugin ($$), or getting very convoluted with multiple side channels working with filtered companding, the simple answer is: you CAN'T.

Scott

[netmask56]

If you go to Effects -> Stereo Imagery -> Centre Channel Extractor -> drop down menu -> Vocal Remover. Might work as long as the track is stereo and there is little cross feed you should be able to get rid of >90% and with a bit of fiddling get what you are after..... lots of varaibles along the way... Good luck

[lordsmurf]

There's a special simple one-click plugin that not many people know about. It's an industry secret. It will remove vocals perfectly, like they're not even there. It's called... Oh wait. Nevermind. It's a secret.

Seriously, this task requires frequency carving. You'll likely carve up the remaining audio as well. Such things are only meant for forensic work, not enjoyable audio. Removing vocals is usually guaranteed to screw up the non-vocal audio as well. The remaining audio will almost never sound natural.

Adobe Audition is laughably pathetic audio software. Even Sound Forge, far more powerful, will have difficulty at this exact task.

[netmask56]

Depends a lot on source material. I have a few vinyl discs left in my collection from the '60's and '70's that I have removed vocals almost perfectly destined to become family karaoke tracks. It's also been done on a few documentaries where (copyright clearance obtained) the vocalist fades out leaving the orchestra playing on and a new V/O An example is the Warner's vinyl release Harper's Bizarre where the centre material can be totally removed leaving just the accompaniment

[Veda]

It is being said uncompressed lossless wave files are best for channel splitting as mono waves & remixing for surround or other purpose. But in my uploaded wave file (just 1 minute sample loaded) unable to take out vocals & music separately.

My requirement want to split it as mono one with vocal dominating & music playing very feeble & one more with music dominating & vocals very feeble. Is it possible?

I have uploaded sample since I'm unable to remove or take out vocals separately.

If it can't be done what type of wave file (specification) should I use to do such job?

[Cornucopia]

Look, this is getting ridiculous. There is no magic wand.

One can separate 2 different signals from a combined signal to a greater or lesser degree based one (or more) of these criteria:

1. Channel/track separation - if Vox is in L and Music in R, it's easy, just extract the channel you need. Similarly with center/surround channels.
2. Inter-channel phase (polarity) differences - If Vox is in L+R in mono, and Music is in L and R in stereo, a Difference mix (L-R) will get you some (weak) Music and NO Vox. A Summed mono mix (L+R) may make the Vox stronger enough than the Music to make that worthwhile.
3. Wide differences in dynamics - if you have Vox as strong signal, Music as weaker signal, you can make Music even weaker still by using a compressor/expander/noisegate to drop the Music further when Vox signal isn't strong.
4. Differences in Pitch - if you have Vox in midrange but Music in Hi + Low (and maybe some midrange), you can use complementary EQ filters (bandpass, bandstop) to isolate the portions that are closest to the freq. range of that signal.
5. Differences in Timbre/Formant/Harmonics - if the waveforms have distinct differences in character, with Izotope and a few other tools, you can "match" and isolate that signal character from the rest. AI might, in the future, get even better at this.
6. Differences in timing - if Voice plays then Music plays, etc, simple editing can isolate the segments you want.

These can often be combined to further improve the isolation. And all the stuff that has been talked about so far, and in other threads, are encompassed in this listing. Note that this really isn't much affected by the toolset, just which algorithm, which settings. Some/most of these things can be done with even basic editors, they just might take longer.


HOWEVER, this clip is in MONO, which rules out both #1 and #2. It has in most places the voice coinciding with the music, which rules out much of #6. When they coincide, both of them are vying for the foreground, so there isn't enough dynamics differences for a compander/gate to follow the CORRECT signal, ruling out #3. The pitch of the voice is smack in the middle of the music's range and the music recording quality is such that there isn't much beyond either, so that rules out #4. That leaves #5, which is the hardest/most complex of the bunch, and there are many features of the voice and the music that are similar, so that pretty much rules it out as well.

The difficulty is not merely in its file format (though it would be ALWAYS better to using a high quality, uncompressed file format as a basis to work on), but on its provenance (how many generations?) and its original design. I would say you are SOL.

If this were a professional project, I would put effort into investigating alternate copies that had only 1 of the elements (in a sense, pre-isolated), or even in re-performing it. As I have mentioned before, the amount of time to get something like this "restored" (enhanced) takes nearly as much effort (in time/money/effort/learning/research) as doing it from scratch.


Scott

[Veda]

Thank. I was also of the opinion it can't be split or taken out. You have explained out with reason well.