I would like to remove the music from an audio (.wav) file.  The audio
file contains both voice and background music and they overlap (it is
a clip from a movie with actors talking and a musical score playing at
the same time).  I need to remove just the background music, leaving
the voice part in-tact.  I would appreciate knowing the names of some
programs that can automatically remove music from an audio file
(leaving just the voice).  Also, any turorials that go through the
process of doing this manually using
such programs as Sound Forge, Cool Edit Pro, whatever.  Basically,
I'll
appreciate any means of doing this as long as it works adaquately and
does not involve expensive audio equipment.  It must be possible to do
this somehow.

---

Clarification of Question by bmhoge-ga on 12 Aug 2002 19:37 PDT

Yeah, this is from a VHS tape.

Man, I'd love to find this too.  I've tried many times, without
success, to do this.  You always end up with some music under the
voices, even with multiple notch frequency filtering.  You can't cut
all the frequencies out - you have to pass the voice frequencies at
least.

Anyone?  Tips?  Ideas?

I tend to agree that this can't be done - at least done well.

However, there used to be a device that did just the opposite called
the Thompson Vocal Eliminator. It purpose was to remove the vocal so
that you could sing along with the music that was left. I also
couldn't imagine that working very well, but they seem to have sold
for many years, so they might have. Perhaps there would be a way to
use it in reverse?

I tend to recall seeing ads for these in Popular mechanics or Popular
Science in the classifieds display section.

I just checked and they are still around! http://www.ltsound.com/ I
have email into them to see if it can be used in reverse and will post
as an answer if I find out it can. Stay tuned.

-=clouseau-ga=-

I do, in fact, believe that this is quite possible to be done.  I
believe the music information is layered under the vocal section and
can be removed using signal processing.  However, I do not believe
that this can be done cheaply.  I remember having to perform when I
was very, very young, to a song that was played on an expensive piece
of equipment which, quite successfully, removed the voice part of the
song and left the underlying music...it may have been luck, but the
music still sounded good.  You see, it is the same deal as why a
computer can take a 2-D scan of a page full of thumbprints, all
overlapped, and can lift a single fingerprint from the bunch.  Or why
most true 3-D (think Star Trek) holograms are printed on a 2-D piece
of film that looks nothing like the hologram itself, but is a mixture
of random ink blotches containing phase information.  It is a matter
of finding the mathematical description and removing and manipulating
what you wish.  If you are interested in the expensive methods for
removing the voice, I will check on that.  I will also ask the Signals
professor at my university if he has any suggestions.
I can hear it now:
"My dear, you should know how to do this.  You sat in on my class many
a time and *sheesh*..."

:)  By any means, I will check!

Dexter

well seeing as how it was most likely made from a video-tape, you
can't do it thru a program, because the audio wav will only see one
track, if you have an original  beta tape of the movie there are
hardwares to seperate those, but I doubt it is, more than likely it's
a vhs tape, and that I dunno of anything affordable to seperate the
tracks, sorry :/ wish it were possible though, DVD there is likely
stuff to do this if you have a dvd rom, maybe go that route, good luck
and let us know what ya find out,"I am a sound-tech-head,my sounds go
into numerous ears therefore I am"

Alas, the folks at Thompson Vocal Eliminator replied that their device
does not do the reverse, it will only eliminate the vocal. It turns
out, it is also $2400!

Does anyone else think of Speakeasys, ballad singers and machine guns
when they read "Thompson Vocal Eliminator"? Can't get that image out
of my mind 8^)

-=clouseau-ga=-

Well, it doesn't really matter if the TVE runs in reverse or not --
you could just take the "vocal eliminated" track and subtract it from
the original track. However, I suspect that you may still be left with
some music because the TVE can remove some music as well while still
performing its function.

I wonder if tools exist to do this by hand.

I still think it is because as long as you can hear the music
underneath the voice, it is possible.  How cheap, however, is another
deal.  It does not matter if it is on VHS or not, as the sound of the
human voice has a distinct frequency over music.  For instance, one
can take answering machine tape and remove the voice on top so that
you can only hear the background noise, and vice versa.  I have
e-mailed the EE Signals professor at the university and am awaiting a
response.  I know that we have to equiptment to do it. You might wish
to contact the Electrical Engineering Department of a local university
(private ones are more likely to have the expensive equiptment laying
around!) and you might ask them to do it for you.  As to the method, I
will e-mail you back when he e-mails me.  I would attempt to answer
this myself, but as I am not a PhD on the subject, I would rather
leave it to the experts.

Cheers,
dexterpexter-ga

UPDATE:  He e-mailed me back and asserted that it is possible through
a filter, and redirected me to an expert on the subject.  I will
contact him and update you further once the "expert" clarifies things
for me.  I will keep in mind when I meet with him tomorrow that you
are trying to stay on a budget, and will try to find the least
expensive means of doing so.  If doing it cheaply is not an option, I
may be able to put you in contact with someone at your local
university that might do it for free, or have a student do it for
practice. :)

I will be sure to update you further after I meet with him.

Cheers,
dexterpexter

UPDATE:  I went to meet with him and got caught up at the university,
and he went home before I got a chance to chat with him.  Another
professor mentioned a specific filter that could do this and exactly
what frequency you would have to first eliminate before filtering the
file.  However, caught in the rush, I do not remember the specific
filter name.  I will *hopefully* catch up with the resident expert on
this Monday.  If not, I will try talking to him over e-mail.  Do not
dispair!  I am still thinking about you...

dexterpexter

With modern movies the most reliable way to do this is probably by
getting hold of a DVD of the movie.  Voice is usually all recorded to
the centre speaker's channel - the likelyhood of being able to extract
it is quite high.

izzard,

I agree that this might be the cheapest route.  Any advice that the
Signals expert can offer will most likely be expensive.  I do know
that a filter can be used to do what this user wants, but the expense
in the equipment will likely not satisfy the criteria of finding cheap
solutions.

The cheapest route would be to rent the DVD from Blockbuster (if this
movie is currently made on DVD) or perhaps a cheaper  dealer that
rents DVDs, and use any of the free programs findable online to record
the movie to capture only the voice.  I know that DVD-ripping programs
can do this.  Legality in doing so, however, is open for debate.  It
is true that the sound is highly configuarable on DVDs.

If you have difficulty ripping the DVD, then I can only suggest that
you grab the .wav file and take it to the nearest University and
consult with their Electrical Engineering Department.  I know from
experience that professors and chairmen are all-too-eager to give this
project to their grad students for practice...free labor.  They might
do it for you for free.

Lastly, I must once again warn that splicing music from the sound byte
could be controversial in legal terms as it may be against the
copyright restrictions.  Be careful.

Good luck and I hope to be of more help soon,
dexterpexter