Jump to content

Stem separation in Sonar?


Brandon Bowles

Recommended Posts

I see that stem separation is all the rage at the moment with RipX, Serato Sample, and especially with FL Studio upping the ante recently and implementing it directly into the DAW.  In the comment section of a YouTube vid I was watching on the subject recently, I was surprised to see  someone mention that Bandlab has stem separation also with the Splitter tool.  I have never used Bandlab so I had no idea about this and just checked it out.  It doesn't sound quite as clean as some of the others on the market right now but it's a start.  I'm hoping they are planning to implement this in the upcoming release of Sonar.  That could really turn some heads.  I'm also an Ableton user, and people are quite upset about the omission of stem separation with the recent Ableton 12 announcement.  It's probably the biggest complaint I keep seeing about it.

 

Just a thought.

Edited by Brandon Bowles
Link to comment
Share on other sites

  • Brandon Bowles changed the title to Stem separation in Sonar?

meaning "separate instruments / vocals from a two-track source"? or "export a project as stems"? if the latter, grouping instruments into using busses and then exporting those busses is a quick and easy way to export stems. or using aux tracks to do the same. my own mix template has exports preset for two-track WAV and MP3, and stem buss exports built-in so a quick recall settings, tweak selections, update task and run. 30 seconds to get stems of a mix exported. 

Link to comment
Share on other sites

BandLab does indeed have the Splitter tool, which will split a stereo mix into separate bass, drums, vocals etc.

I believe Cakewalk Next will have this feature too.

I'm not sure if the new Sonar will have this feature on day one of release, but I AFAIK it's in the pipeline.

 

  • Like 4
  • Thanks 1
Link to comment
Share on other sites

I'm guessing that with the advent of AI processing, this will become a common thing soon. -In my mind, socially/creatively it would seem to be a successor to the DJ remixing trends from the 90's & such. Now we will just have to put up with complete individual tracking remixes put together (and rights fought over) by 3rd and extended parties.  Hopefully somehow original content will still remain to be created anew, but for now, a swath of "new" techno remixes of "classic" material "re-imagined" and in Dolby Atmos is on the way!  -Don't get me wrong - I am enjoying some of the newer classic remix material & projects.  -Just some of it will go a long way though, for me anyway!

  • Like 1
Link to comment
Share on other sites

Variants of the AI code can be found on git-hub (i.e. it's free), and I'm pretty sure a bunch of commercial implementations are based on this, only differing in the training data they've used. I've not seen the BandLab code however, so I can't speak for that.

There are some extremely useful applications of this technology, depending on how you've trained it.

For example:
- Being able to split up drums when you've only recorded with 2 or 3 mics
- Removing/reducing bleed from other drums... in fact removing/reducing bleed from any recording.
- Being able to tweak the balance of instruments 
- Splitting chords into separate notes

 I've not personally seen any examples that are specifically trained in drums so far, but it's certainly do-able using the technology. I'd be surprised if it doesn't already exist though.

The only downside is the artefacts that are pretty much always present in the results to some extent.  For remixing, this is normally not an issue as when the stems are re-combined, they blend almost seamlessly.  Tracks in isolation however might need some treatment.

When I've played around with this,  I've found the best results are when you combine the separated stems with the original, using a separated stem to boost/reinforce the original part. Keeping the separated stem at a fairly low level is enough to give that part a volume boost over the original without any artefacts becoming obvious.

 

  • Like 5
Link to comment
Share on other sites

It would be super popular with people that make backing tracks for live performance as well as Karaoke tracks. It would probably be breaking all sorts of copyright rules but if used by individuals and just at local performances doubt if they would ever be called out on it. 
I would never use an original version as a backing track but what comes to mind is if you could break a cover song down into a lot of separate tracks you then could then convert a lot of it to midi. Now you can use that to roll your own version and not breaking the law anymore. 

This would save me a little bit of work. Right now I just drop the original song in a project and re play the parts by playing along with it. It drag the song to the timeline first to create a tempo map. 
So I can see the stem separation working at least to rip the bass part and parts of the drums. You’ll never be able to turn cymbals into midi properly. 
Anyhow yes a useful tool in many ways even for real musicians. 

Link to comment
Share on other sites

8 hours ago, msmcleod said:

When I've played around with this,  I've found the best results are when you combine the separated stems with the original, using a separated stem to boost/reinforce the original part. Keeping the separated stem at a fairly low level is enough to give that part a volume boost over the original without any artefacts becoming obvious.

That makes a lot of sense. I wonder if that's what iZotope is doing with their master rebalance feature. I've used it to emphasize vocals/drums/bass from vinyl cut in the 60s.

Have you tried processing the stems when you combine them with the original part?

Link to comment
Share on other sites

2 hours ago, John Vere said:

you could break a cover song down into a lot of separate tracks you then could then convert a lot of it to midi. Now you can use that to roll your own version and not breaking the law anymore. 

Well.... That's always been rather a gray area in a truly legal sense @John Vere. Be careful, because from what I've seen, it's all a matter of scale (as with anything litigious really). To put it simply - if your use case garners you a lot of money (exposure that leads to that, what have you), then rights holders will come after you. It has always been that way. -And if you were one of the creators, producers, (wait, they are usually last behind) -rights holders of such material, you would look at it that way in most cases as well.

With regards to re-using "cover material" - as long as you stay under a certain financial/exposure level, you will be just fine most of the time. -As to using MIDI representations of copyrighted material, and regarding this original post topic - in using AI or otherwise derived "stems" goes, you won't want to be taking that belief to court. I agree with @OutrageProductions that we will see a lot of that kind of trouble to come from this. And at the same time really, from a positive perspective, the average Cakewalk user should be able to use it as another creative tool, as @msmcleod points out, or yes Johnny V :) -as a great option for creating backing track MIDI & whatnot.

 

As an aside - ever notice how MIDI backing track material is curated? Most of the "free" repositories, versions, are rather sketchy to obtain, and/or just plain bad arrangements that probably are so far off as to avoid major copyright. -There are monetized, official versions, and even services that will provide them for a fee though.  -In a related fashion, you will find that playing cover tunes onstage, at a certain level of the business, -or of course for charity or academic purposes, has always been no big deal - right? But you won't find any major venue, or act on a major, making their business playing "cover" renditions, or having that kind of material without paying rights usage fees.

-Even some of the small places I played in cover bands years ago had issues with performance rights dues, once they were singled out or became prosperous enough. I've had to advise businesses on whether they could use their own tracks for on-hold music, or general in-house piped music systems. -The point is, taking & reusing existing works can be tricky, but hopefully not for the average user. I look forward to using some of the techniques being talked about here, and hopefully none of us will get tripped up by any of the legal wrangling that will inevitably be involved with "ripping stems from a mixed track", or whatever it gets labeled as.  And hopefully, just like with cover renditions, some new & interesting creativity will be involved, as well as continued original material - without AI - maybe?😅

  • Like 1
Link to comment
Share on other sites

Yep. No danger of me monetizing a cover tune.  I’m even careful as I make midi tutorials about what I play in the demo. I mostly choose my originals. I have found that the robot detectives don’t seem to spot midi generate backing tracks. They are as you say just way to far off base. 
And even Dentist offices have to pay for Muzack. My wife worked most of her career in them. In the past you had to pay for a service and it was expensive . Using your own CDs or even the radio was not allowed. 
Now all they need to do is pay for a Spotify account. 
I played at an annual local event for years and out of the blue they got a warning from Socan that they need to pay. I told them to tell Socan all the bands  only play original music. They never heard a word back. 
Im almost done with cover songs anyway. Starting to head into the wonderful land of gigs that want original music.  
 

Link to comment
Share on other sites

Yeah, rip it into stems, copy it, steal it, use until you're blue in the face - if you are the average broke musician. -Heck, with only 12 major tones in the western scale, you can accidentally happen upon a melody like one produced & sold before somewhere - mathematically, it's not unrealistic, by just chance. -Humming a tune from your subconscious, maybe listening to bird song (who is looking out for their IP..?)  music seems so simple. -But if you have lots of money involved, that becomes a whole other thing in the western world.  Just ask the spirit of sadly departed George Harrison, and on & so forth.🙊😑

  • Like 1
Link to comment
Share on other sites

On 11/18/2023 at 8:18 PM, JnTuneTech said:

Yeah, rip it into stems, copy it, steal it, use until you're blue in the face - if you are the average broke musician. -Heck, with only 12 major tones in the western scale, you can accidentally happen upon a melody like one produced & sold before somewhere - mathematically, it's not unrealistic, by just chance. -Humming a tune from your subconscious, maybe listening to bird song (who is looking out for their IP..?)  music seems so simple. -But if you have lots of money involved, that becomes a whole other thing in the western world.  Just ask the spirit of sadly departed George Harrison, and on & so forth.🙊😑

doesn't matter. with AI, all you song belong us now.

Link to comment
Share on other sites

Just an update. I installed Next and been trying it out. I took a master of one of my songs and dropped it in and ran the track separating tool. I goes to the server and comes back. You get 4 tracks. Bass, Drums, Other Instruments and vocals. 
There was noticeable distortion on the bass track when soloed. The drums sound good, the vocals sound ok but reverb tails are gated. Everything else ends up in the other instruments track which includes the missing vocal reverb tails. It seems to sound fine when you play all 4 tracks.  
But here’s something cool I thought of.
I have a lot of old master wave files from my 4 and 8 track days where the multi track is long gone. Some are good songs but the mix was not great. 
I dropped one of these into Next and once I had the 4 tracks I turned up the bass and vocals and now the song sounds 100% better.  
So another tool that you can use for remixes. 

  • Like 3
Link to comment
Share on other sites

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
×
×
  • Create New...