Eliminating vocal parts of a song
Bill Olivas
I am interested in a software package that can eliminate the vocal
parts on a given song. Is this possible? It doesn't sound too hard to me.
Also, is there a package that can translate sound to sheet music?
What I'm trying to do is this:
I happen to like a particular song on a cd I recently bought.
This song most likely won't be released as a single since it doesn't have
much
commercial appeal to it. After talking to a local sheet music store,
they informed me that if it isn't in fact released as a single, the only
way for
me to get the sheet music to it would be if the album songs are released
in a book form. I'm thinking that this is improbable.
Anyway, the song is backed musically by an acoustic guitar. So if I could
remove the vocals I would then have a track with only a guitar playing.
Shouldn't it then be pretty simple for a software package to then spit out
some sheet music? I would think so but I knew anything I wouldn't be here
asking questions.
If enough details weren't given feel free to let me know and thanks in
advance.
Bill
--
**************************************************************************
Bill Olivas aka "can I see your license and registration, punk"
Apple Computer
oli...@newton.apple.com
**************************************************************************
John Churchill
oli...@newton.apple.com writes:
>Anyway, the song is backed musically by an acoustic guitar. So if I
could
>remove the vocals I would then have a track with only a guitar playing.
>Shouldn't it then be pretty simple for a software package to then spit
out
>some sheet music? I would think so but I knew anything I wouldn't be
here
>asking questions.
I don't think that what you want to do is possible once the mix is done
and the CD pressed. I have experimented with soundeditpro but to no
avail, I can't even slightly mask out any particular instrument, much
less the vocals.
Thomas M Gooding
The only way that it can be done is if you get a recording of just the voice
and subtract the waveforms. The voice also must be at the same amplitude as
the voice on the track. Otherwise, I think its impossible.
--
Tom Gooding
tgoo...@iastate.edu
Iowa State University
Scott A Alburger
: >I don't think that what you want to do is possible once the mix is done
: >and the CD pressed. I have experimented with soundeditpro but to no
: >avail, I can't even slightly mask out any particular instrument, much
: >less the vocals.
: The only way that it can be done is if you get a recording of just the voice
: and subtract the waveforms. The voice also must be at the same amplitude as
: the voice on the track. Otherwise, I think its impossible.
Removing the vocal from audio tracks *is* possible. There is a stand-
alone product called The Vocal Eliminator that claims fair results. It is
manufactured by: LT Sound
Dept. MX-9, 7980
LT Parkway
Lithonia, GA 30058
(404) 482-4724
(404) 482-2485 24hr phone demo line
The company has been around a while and sends out free demos. According to
the demo they sent me a while ago, it is successful only about 50% or so of
the time. Rather pricey too. The company advertises in audio trade magazines
like MIX and MUSICIAN.
There *should* be a Mac solution for this application (some of the higher
end cards like Digidesign ProTools may have limited applicability) At this
point, it seems the demand has been too low and legal ramifications too
high for such a product to be widely available (removing the vocals from a
song suggests repackaging & copyright violations to the music industry)
Sad but true.
Just for the legal record as well, I have no connection to either
Digidesign or LT Sound.
Good luck-
S.Alburger
Multimedia Specialist
Wharton Academic Technology Services
University of Pennsylvania
ALBU...@wharton.upenn.edu
Fred Schaerlig
the Beatles voice, but try to run it on something more recent, in which most
of the voice has been first sampled before beeing played back on the tune...
you'll get surprises...
Fred
---
Fred Schaerlig / ArcAngels
fsch...@di.efpl.ch / FIDO: 2:301/811.16
All probabilities are 50%. Either a thing will happen or it won't.
------------------------------------------------------------------------
Martin Dupras
The topic of vocal or instrument removal from a recording has been a long-time
preoccupation, and part of this is because it's *much* less trivial than it
would seem in the first place...
In short, the problem is that it's pretty difficult to recognize in a signal
which partials belong to which instrument. As of now (to my knowledge, at
least) state-of-the-art technologies permit full removal or extraction of
vocals from a recording, but then you *definitely* lose much quality.
Some "shortcut" techniques allow vocal removal (as with the so-called "vocal
eliminator") in some cases, provided that
1) the voice is mixed exactly dead-center
2) the other instruments are not
For example, this will work very well with some old beatles recordings. This
technique is pretty easy (try it with soundedit or sound designer II!). All you have to do is invert the phase of one of the two channels, thus canceling out what was in the middle. As you'll notice, you'll definitely lose some quality
anyway.
Now, probably the best but still not-entirely-succesful way would be to analyze
in a signal what the fundamental frequency of a vocal track is, determine the
"most-probable" amplitude of partials, and thru some fast-fourier transform
(fft in short) analysis-synthesis technique (McAulay-Quatieri, for example)
resynthesize everything on the original *except* the expected vocal components.
Some components from the vocals will probably remain to some extent, but will
probably be masked by other components in the signal.
Other techniques are being developed, so there's probably some more stuff that
I don't know of, but basically that's where it's at...
- Martin
--
Martin Dupras : "She goes channel to channel
dup...@ere.umontreal.ca : Cold lookin' for a hero
musique, univ. de Montreal : She watch channel zero!" - Public Enemy
.......................................................................
Hiro Sugawara
This was one of my studies at a graduate school back more than 10 years
ago. We finally succeeded in producing acceptable levels of
"music-minus-one" sound. It was done by using state-of-the-art digital
signal processing techniques (at the time). The main theory was to
compute self corelation of waveforms to find pitches and amplitudes
of the sounds of interest, based on the philosophy "if the human can
recognize it by some logical means, so can a machine."
It needed a lot of computational power (hours of 16 bit processor time
to produce several seconds of sound). Today, we can use a lot more
powerful computers but it may be still difficult to provide such a
real-time processor at an afordable price.