Isolating Voice only in WAV

[wizaerd]

I have a WAV file that's a recorded song, but if I try to use it for my lip synching in a switch layer, the mouth moves with just the music. How can I best isolate just the voice of a recorded WAV file?

[Lost Marble]

An audio wizard may be able to answer that, but one approach would be to use Papagayo to manually lipsync the voice, rather than relying on the automatic lipsync.

[slowtiger]

Short answer: you can't extract a voice track from a complete recording with music and stuff, not even with the kind of gadgets they seem too use in CSI and NCIS.

Workaround: Just sing the track yourself. Put the original on headphones with low volume, record your own voice only. It doesn't matter if you can't hit any note, it is good enough if you just speak the lyrics in the correct tempo. Now take this recording into Papagayo, and voilá!

PS. Don't forget to replace your voice with the original song for the finished film.

[myles]

slowtiger - nice solution! Creative and practical!


Just thinking aloud:

With some commercially recorded music, they put the voice on both stereo tracks, but different instruments on each track.

By inverting one track (or something like that - I've only heard about this) you can combine the track and the voice cancels out to a large extent, leaving mainly instruments.
It's a technique used by some karaoke software.

Now, I wonder if by somehow inverting that result and combining it with the full track, could you mainly cancel the instruments and leave mainly just the voice?

I have no idea if this is workable or not.

Regards, Myles.

[Bones3D]

Just thinking aloud:

With some commercially recorded music, they put the voice on both stereo tracks, but different instruments on each track.

By inverting one track (or something like that - I've only heard about this) you can combine the track and the voice cancels out to a large extent, leaving mainly instruments.
It's a technique used by some karaoke software.

Now, I wonder if by somehow inverting that result and combining it with the full track, could you mainly cancel the instruments and leave mainly just the voice?

I have no idea if this is workable or not.

Regards, Myles.

This is actually not accurate. Wave inversion is generally used for noise cancelation purposes... and only works on the "noise" in question in its entirety. Basically, it'd competely eliminate the entire sound (music and vocals) in one fell swoop.

Many of these pseudo-karaoke devices/software instead rely on a "one-size-fits-all" form of frequency filtering to "remove" the vocals from a given song. Generally speaking though, the vocals are still present in the sound, but get muffled out enough to be overpowered by the music. Also, parts of the music itself usually get muffled out well.

While this does suggest that doing the reverse to filter out the music would probably work, the end result would probably not be worth the time and effort involved to do it with any sort of intelligence behind the process.

I imagine if you ever want to explore this for yourself, software like ProTools and Logic probably have filters that can be used to do this sort of thing.

[heyvern]

Another... really really long shot option...

If the song has been used in a movie... rent the DVD... sometimes when they mix the audio for surround sound... the singer MIGHT be "extractable".

You would have to have some kind of surround sound system... and a way to isolate it.

This.. is... a long shot... a real long shot.... one in a million.

-vern

[myles]

Hi Bones3D,

This is actually not accurate. Wave inversion is generally used for noise cancelation purposes... and only works on the "noise" in question in its entirety. Basically, it'd competely eliminate the entire sound (music and vocals) in one fell swoop.

Ah, but this trick relies on the fact that the only identical content on both tracks is the voice - the instruments on left and righ channels are different. So, the left track contains the voice and, say, drums and guitar, while the right track also contains the voice and, say, electric organ and trumpet.
Of course, not all music+voice is recorded this way, but it will work with some.

Splitting the 2 stereo tracks into 2 mono tracks, inverting one, then combining them, only cancels where the 2 tracks were identical - this will get mostly the voice, although it doesn't work 100% because voice and instruments are mixed.

There's a tutorial for doing it in Audacity here: https://wiki.brown.edu/confluence/displ ... y+Tutorial

Regards, Myles.

[DK]

Get a microphone and reord yourself singing along with the song then use that audio file as your auto lipsinc master.

D.K

[DK]

Oops...sorry did'nt see slowtigers post.

D.K