remove music, keep vocals

piet piel · Post by **piet piel** » Mon Jun 28, 2004 4:32 am

Hello!

Anyone knows how to remove (most of the) music from a song and keep only the vocals? Removing vocals is easy, this could be harder, right?

Regards,
Manoah

[ This Message was edited by: piet piel on 2004-06-28 05:33 ]

Counterparts · Post by **Counterparts** » Mon Jun 28, 2004 6:29 am

piet piel wrote:

Anyone knows how to remove (most of the) music from a song and keep only the vocals? Removing vocals is easy, this could be harder, right?

Regards,
Manoah

(Ponders...)

What about if you remove the vocals, then phase-invert the resulting audio and mix that with the original track? That should cancel everything except the vocals, no?

Royston (probably talking rubbish)

piet piel · Post by **piet piel** » Mon Jun 28, 2004 4:17 pm

Yes, good thinking!! Thanx!

But I already did that

and it didn't work. I dont know why.. perhaps i didn't do it right. Theoretically, if you have a 180 degrees stereofield, why is it not possible to cut everything just next to the mid? It sounds so logical! But perhaps that's tomorrow's technology...

[ This Message was edited by: piet piel on 2004-06-28 17:19 ]

astroman · Post by **astroman** » Wed Jun 30, 2004 6:12 pm

dunno exactly how a vocal remover operates, but cancelling out everything except the center is probably only one part of the process - and so the 2nd inversion fails.

The most significant parts except the location are frequencies and transients a 'regular' human voice cannot (or isn't expected to) perform, like hi-hats, kick, snare, bass.

I assume they transform the original signal to do an analysis of those parts and then everything 'suspective' is removed.

The reverse process would be much more difficult, as the human voice has many components which are similiar to instruments or certain parts of their playing range.

A very interesting question from the engineering point of view - and certainly one where 'traditional' quantitative analysis (which bands, how fast) will fail.

It resembles in fact a description of future technology as the change of sound over time must be reverse processed to determine the origin of the various sources to be able to split them up.
It reminds me a bit on how the Hartmann Neuron synth codes it's sound models.
They apply artificial neural networks to do the analysis part of the job.
It's no new technology though - I have a bunch of books about the subject reaching back to 1988

cheers, Tom

ChrisWerner · Post by **ChrisWerner** » Sat Jul 03, 2004 6:59 am

One method to separate voice out of music is the fingerprint method. Like the Waves x-noise. It works very well to remove noise but it has also presets to work on vocals remove.
X-Noise "learn" a song and creates a profil of it. Once the profil is generated you can try to separate noise, vocs etc.
You can hear the music without vocals or you can switch to hear the bits that are cut away, the difference.
I´ve tried X-Noise to separate vocs out of music, I test it with the musical version of War Of The Worlds.
The result is bad, to many instruments use the same freqs where the voice lies. The vocs sounds like passing a filter and many artifacts from the music around it, too.

Tom, the thought on how Neuron analyses the bits are very inspiring, I wish I had one Neuron in my studio.

[ This Message was edited by: ChrisWerner on 2004-07-03 08:02 ]

siriusbliss · Post by **siriusbliss** » Fri Oct 08, 2004 9:37 pm

I've been seeking a way to do this for several months now on a track I recorded during a jam session.

I've tried the remove vocal/phase inversion method. I've tried EQ'ing. I've checked out 3rd party programs. Nothing has really worked.

Seems pretty complicated, but there's bound to be someone out there working on this.

dodge77 · Post by **dodge77** » Sun Nov 07, 2004 12:05 pm

Have you tried a vocoder such as the "Orange Vocoder" from PROSONIQ? Maybe you can use it to (partly) isolate the vocals without patching it through the synthesis section. It all depends of what you want to do with it after isolating it of course...

Dodge

Post by **garyb** » Thu Nov 11, 2004 11:16 pm

pretty much impossible. the music and vocals are just one big file. if you remove one element, the resulting mess may still be listenable because there are many things happening and only SOME of the music you want gets removed. doing the opposite only shows how pitifully bad an algorhythm is at differentiating between one sound and another. all 1s and 0s look alike.....

Daniel Mart · Post by **Daniel Mart** » Sun Mar 05, 2006 8:50 pm

I used Fleximusic Wave Editor using Band Pass Stop Filter in that you can either do blocks or keeps the frequency range. But you have to find by trial and error in which frequency bands to pass/stop.