splitting up instruments in a track

[Jack Hawk]

I am working on a new project that requires taking a pre existing track with a rhythm and drum track, and pulling the guitar track from it. I am wondering if there is any software available that can do that, isolate a lead guitar and remove it  or isolate it by itself. The final track is intended for a tribute song to a deceased friend. Any thoughts on whether this can be accomplished and what I'd need to do so? Many thanks!!!

[OutrageProductions]

You could try email. Contact the original composer, publisher, or record label and request a copy of the original multitrack.

Even the techniques to isolate or remove vocals are only minimally successful, so a lead guitar? Extremely doubtful. 

Sound waveforms are like paint... once you combine red & blue pigments to mix into purple, the convolution is nigh impossible to undo.

Edited April 19, 2023 by OutrageProductions

[sadicus]

You could try this RIP-X:  https://hitnmix.com/

I had some old stereo studio mixes  (drum, Bass, LR Gunter, 2 vox, sample FX) and it worked good enough.

[abacab]

Try iZotope RX 9. The "Music Rebalance" feature can be adjusted to filter out 4 stems with automatic stem separation to generate individual files: Vocal, Bass, Percussion, & Other. Or just rebalance the mix, for example remove the vocal or isolate the vocal.

YMMV with results, and of course it always depends on the source material.

https://www.izotope.com/en/products/rx/features/music-rebalance.html

I believe that Steinberg SpectraLayers has something similar.

 

Edited April 19, 2023 by abacab

[Jack Hawk]

Thank you all for the suggestions. OutrageProductions, I am trying but have been unsuccessful so far and I want to have a backup plan inplace. Sadicus and abacab, thank you,I will give the suggested software a try. I own an older version of the RX software and will see if mine has that functionality. If not, I can run a trial and see what happens. I appreciate the  suggestions.

[OutrageProductions]

13 minutes ago, Jack Hawk said:

 I am trying but have been unsuccessful so far...

I know that getting anything that has a currently effectual copyright released, even if only for educational purposes, is like pulling teeth without anesthesia, so I  feel ya. Good luck.

[mettelus]

Just to clarify, are there two guitars on that track (rhythm and lead)? If so, that is the primary hurdle due to frequency overlap, but time-based FX (delays, reverbs, etc.) tend to cause the most grief trying to separate things. If you hit a wall with it and want another to take a shot at it, shoot me a PM. I will not have time to sit until this weekend though.

[Jack Hawk]

6 hours ago, mettelus said:

Just to clarify, are there two guitars on that track (rhythm and lead)? If so, that is the primary hurdle due to frequency overlap, but time-based FX (delays, reverbs, etc.) tend to cause the most grief trying to separate things. If you hit a wall with it and want another to take a shot at it, shoot me a PM. I will not have time to sit until this weekend though.

Thank you mettelus, I will let you know. I havea few ongoing projects myself and probably won't get to try this until the weekend myself. I appreciate the offer. 

[Bruno de Souza Lino]

On 4/19/2023 at 3:48 PM, abacab said:

Try iZotope RX 9. The "Music Rebalance" feature can be adjusted to filter out 4 stems with automatic stem separation to generate individual files: Vocal, Bass, Percussion, & Other. Or just rebalance the mix, for example remove the vocal or isolate the vocal.

YMMV with results, and of course it always depends on the source material.

https://www.izotope.com/en/products/rx/features/music-rebalance.html

I believe that Steinberg SpectraLayers has something similar.

 

SpectraLayers, RX, Spleeter, Acustica and other similar products use the same MusDB dataset, so you'll get similar results regardless of which one you use. Demucs V4 Hybrid tends to produce better separation and was what I used to do the two mashups I posted a few days ago to the forums. The output still isn't perfect and you'll be missing quite a bit of content above 11k as it still uses the same dataset with some 800 more songs added, but it's a different algorithm.

[Bapu]

Talk to Peter Jackson, he claims to have superior AI for that sort of thing.  😉