Phase inversion "Trick" for isolating vocal not working.

brassplyer

brassplyer

Well-known member
Several times I've successfully peeled the track away from a vocal by lining up their recording with the original backing track then phase inverting just the backing track. It's even worked if their recording was a mono mixdown leaving just a trace of the backing track portion of their recording. Why I would do this is re-mixing a recording where there's an issue and don't have access to the isolated vocal otherwise.

I find some really obvious peak that stands out in both - the waveform generally isn't absolutely identical but close enough where the peak is clearly defined. Then zoom in tight and adjust one track back and forth sample by sample to get them as closely lined up as possible and find the spot and volume balance between the tracks where the backing track becomes the least audible.

However I tried this using Cakewalk by Bandlab and I can't get it to work. Any theories as to why this might be?
 
If there's an overall eq (or perhaps an all-pass filter) on one that's not on the other, that could rotate the phase in interesting ways at different frequencies making it essentially impossible to cancel the signal at all frequencies.
 
Spectral layers is really good for rebalancing. In the usual mode, you split out the vocals, bass, drums, piano and the rest to five faders, the you can rebalance and export it. It works really well. You can remove any of these things totally if you want, or remove others exposing tracks. If you’re a drummer, you could replace the drums pretty well with new ones, the same with the vocals or piano. What gets put into ‘others’ Is a bit variable, so it can be the strings, or synths, or even the BVs. It’s not cheap but effective.
 
bouldersoundguy said:
If there's an overall eq (or perhaps an all-pass filter) on one that's not on the other, that could rotate the phase in interesting ways at different frequencies making it essentially impossible to cancel the signal at all frequencies.
Click to expand...
On this particular one I doubt it since it's just a karaoke performance and I'm using the same karaoke track. Granted mine is a *recording* of the track but that's always been fine in the past.
 
rob aylestone said:
Spectral layers is really good for rebalancing. In the usual mode, you split out the vocals, bass, drums, piano and the rest to five faders, the you can rebalance and export it. It works really well. You can remove any of these things totally if you want, or remove others exposing tracks. If you’re a drummer, you could replace the drums pretty well with new ones, the same with the vocals or piano. What gets put into ‘others’ Is a bit variable, so it can be the strings, or synths, or even the BVs. It’s not cheap but effective.
Click to expand...
Is there something similar that works with Cakewalk?
 
The trouble is they're not really the same thing - they're pretty good for extracting audio - producing the so called 'isolated' tracks, or karaoke tracks but they rarely do it without obvious artefacts.
 
They're running at different speeds. If you line it up in one spot, it quickly drifts so it's misaligned in a very short time. It might be possible to stretch or squeeze one to match the other (if the speed difference is consistent), but I haven't tried that yet.

[Edit] It's not consistent. It's like one was recorded on a cassette deck with speed that drifts up and down.
 
Last edited:
bouldersoundguy said:
They're running at different speeds. If you line it up in one spot, it quickly drifts so it's misaligned in a very short time. It might be possible to stretch or squeeze one to match the other (if the speed difference is consistent), but I haven't tried that yet.

[Edit] It's not consistent. It's like one was recorded on a cassette deck with speed that drifts up and down.
Click to expand...
Interesting. In both cases the backing track pitch is lowered via a function on the karaoke site, which may introduce inconsistent variations? They were both recorded with a "What U Hear" style function off the sound bus with Audacity.

This is the first time I've tried it on this particular machine where I'm using the onboard sound. When I've done it in the past it was using a machine with an M-Audio Audiophile 2496 sound card. I don't know if that could make a difference.

Is there such a thing as a plugin that will auto-align them?

Thanks for the input!
 
Yep, a pitch changer was on my list of suspects. It looks like it did the process a bit differently on both tracks, so I don't think there's any way to make them line up well enough for canceling to work. A spectral processor might be your best bet.
 
You must log in or register to reply here.
Back
Top