Separating distinct vocals in RipX?

[somethingcleveridunno]

Been playing with the program to make some stems. In the screenshot, Crosby, Stills and Nash's vocals all look very separate, but they're included in the same stem. Any way I can break those out separately? I really really want to have harmonies broken out into different channels.

 

[zeerround]

If the different vocal parts are panned differently in the original stereo, you can use a pan based upmixer like specweb or specscript on the vocal stem. e.g. the main vocal will be in C and the harmonies in either the rears or fronts. Then you can remix those as you like.

Short of that (and manual spectral edit), so far Demix Pro is the only stem generator I am aware of that is separating vocal parts, but Its only a matter of time before someone does the training and freely releases an AI model for one of the opensource tools.

 

[somethingcleveridunno]

No panning on the vocals, unfortunately. Man, that stinks. The software's indicating it sees those as three different vocals, with those three distinct lines.

It's a pretty clear and nice stem, but all three voices layers over each other when I really want to have the lead in the center, and put the two harmony parts in the surrounds for a 5.1 mix. Kinda kills it as an upmix candidate if I can't separate those.

EDIT: Though Demix Pro isn't web-based, so... I could try running the stem thru that... hmm...

 

[quadsearcher]

Cool thing about CSN (well, not so cool for upmixing) is that they sang around one mic (I don't know if that's true for all their songs, but at least some from what I"ve read). The sound of their voices was blending in the air before it even hit the microphone. I know from limited first-hand experience that this can be a thrilling way to harmonize. It seems to create extra harmonics (or "magic") that wouldn't exist from separate mics isolated.

 

[somethingcleveridunno]

Cool thing about CSN (well, not so cool for upmixing) is that they sang around one mic (I don't know if that's true for all their songs, but at least some from what I"ve read). The sound of their voices was blending in the air before it even hit the microphone. I know from limited first-hand experience that this can be a thrilling way to harmonize. It seems to create extra harmonics (or "magic") that wouldn't exist from separate mics isolated.

Ah, yeah, I've heard a couple recording engineers/producers say stuff to that effect, and that does make sense I guess, that the mic would react differently to what you'd get recording them each separately.

And yet, for surround... I want David in the left surround and Graham in the right surround and Stephen in the center and I want to be swallowed in that while still being able to pick each one out.

 

[ar surround]

If the different vocal parts are panned differently in the original stereo, you can use a pan based upmixer like specweb or specscript on the vocal stem. e.g. the main vocal will be in C and the harmonies in either the rears or fronts. Then you can remix those as you like.

Short of that (and manual spectral edit), so far Demix Pro is the only stem generator I am aware of that is separating vocal parts, but Its only a matter of time before someone does the training and freely releases an AI model for one of the opensource tools.

Even DeMix Pro has problems separating the CS&N (and Young) voices. The voices are all there, but they bounce around from stem to stem.
I had the program perform multiple voice separations on the song Carry On. When the group sings, "Carry on, love is coming, love is coming to us all," Graham 'walks' across the front sound stage from the left channel to the right channel. Very cool, but very weird.

And yet, for surround... I want David in the left surround and Graham in the right surround and Stephen in the center and I want to be swallowed in that while still being able to pick each one out.

I did a separation on Southern Cross with DeMix Pro. I got Stephen pretty much in the center and Graham / David in the surrounds. But splitting Graham and David was not successful. Much of the vocals on Southern Cross is panned across the fronts, so it is easy to do a separation of the center from the left and right with a DAW; but whether this method will isolate Stephen from the other singers is another story.

 

[fripppp]

Any way I can break those out separately? I really really want to have harmonies broken out into different channels.

Not an AI tool, and quite old as a matter of fact, but I believe that Melodyne might be able to do something to this effect, and you can probably get something even cleaner if you feed it the AI vocal track instead of the original audio as I did. I was messing with it before AI separation was around and I thought it was pretty good at separating out harmonized vocals.. But I think you have to manually select what notes belong to which singer to separate them, so it might get a bit tedious to do with multiple songs. It's paid software, but I believe there should be a trial somewhere.

 

[mandrix]

Spectral edit seems like your best shot to try.
For example you can delete everything but the notes in the editor that contains the voice you want to isolate.
It will still be a crap shoot if all were singing into one mic though, and spectral editing can be very time consuming.
I don't have vast experience with the spectral editor in RipX but I have done a few edits.