Feature Request: Stem Separation in Sonar

[Jim Stamper]

I can't be the only one who'd like to see Stem Separation included in Sonar, even the paid version.

However I haven't found any other requests for it it.  Or I haven't found the right forum to request features.

It would be great to have this tool integrated with the DAW where I do all my work. I'm not interested in the other features of Sonar Next, and I'd rather not install Next for just 1 feature, which wouldn't be integrated with my workflow.

Thanks

[Glenn Stanton]

i use the stem separation in Next (it basically uploads the file to the processing cloud which does the work, then downloads it) selecting "add stem tracks" adds the separated parts (drums, bass, vocals, other = guitars, keys, etc) to the Next project.

Next seem stable, easy to use, minimal footprint etc. i don't really use it except to do the separations and since i can output stems as WAV files when i'm done, makes it easy to bring into Melodyne for corrections and MIDI conversion or into Sonar as audio tracks. 

since i do a lot of workflow stuff outside of the DAW (composition, initial project structure (Hookpad -> MIDI), generate audio (standalone apps  mostly), and so on, the DAW (for me) is the raw recording of audio tracks (which i then export as WAV and using RX and Acoustica for prep work), and mixing and mastering only. Next is just another tool to get things done.

maybe they'll add the stem separation to Sonar as it seems like a fairly separated piece of code and processing as it stands.

[Starship Krupa]

11 hours ago, Glenn Stanton said:

maybe they'll add the stem separation to Sonar as it seems like a fairly separated piece of code and processing as it stands

As you say, the actual separation is done by an engine on BandLab's server. It's probably the same stem separation engine as the BandLab DAW's use.

Seems like Sonar could someday get whatever mechanism uploads the song and then downloads the separated stems.

I did a test of stem separation among various DAW's I have that include the feature and the Next/BandLab did the best job, IIRC. I use stem separation for learning songs, doing my own version, etc.

[Ivica Ivaskovic]

+1 for stem separation. 

Almost all major daws have it. Some of them, Logic for example, after the last update provides an incredible quality of separations because of one of the latest models (BS Roformer SW - also available on the MVSEP site for free) used.  So, no ancient history htdemucs or spleeter models just to say it's implemented. Do it right. It is cutting edge at the moment and should be treated as such. Accordingly, all other benefits offered by AI would have to find their way into Sonar if there is any desire for the product to survive on the market this time.
And speaking of AI, It's almost incomprehensible that in 2025 I have to spend almost an hour of torture trying to extract a tempo (via audio snap, melodyne simply does not achieve a satisfactory result) from a prerecorded source (paradoxically, the track contains only a metronome in CD quality) only to end up with the message "it cant blah blah blah ... out of range".  Ughhhh.  I am on the last atoms of support for this software experiment. In the free UAD Luna AI does it with one click of a button.

[Bass Guitar]

How I sometimes work is I load a song into Next. Run the stem separation tool and then I use the Export as a Cakewalk Interchange Format CXF file. 
 

Now I just open this in Sonar. 
 

But!! At this point you will be seeing that the down side of the Bandlab system is the Wave Files are 44.1/ 16. I only like to use 48/24 minimum. Not that it matters for stems as they are terrible quality still anyway. I only use them as a reference to recreate the original song. 
 

So for me Audacity and its new stem separation tool is preferable because you can choose any format. It also works locally using your CPU. 
 

[Mark Morgon-Shaw]

On 7/24/2025 at 7:44 AM, Ivica Ivaskovic said:

And speaking of AI, It's almost incomprehensible that in 2025 I have to spend almost an hour of torture trying to extract a tempo (via audio snap, melodyne simply does not achieve a satisfactory result) from a prerecorded source (paradoxically, the track contains only a metronome in CD quality) only to end up with the message "it cant blah blah blah ... out of range".  Ughhhh.  I am on the last atoms of support for this software experiment. In the free UAD Luna AI does it with one click of a button.

There's some good use cases for AI, stem separation and stuff like AI tempo maps are the type of new tools that enhance music production rather than replace it.

Maybe the Luna team is bigger I dunno  - I remember when Sonar used to be the first to implement new features, now it feels like they will never catch up. 

 

[Amberwolf]

On 7/23/2025 at 11:44 PM, Ivica Ivaskovic said:

incredible quality of separations because of one of the latest models (BS Roformer SW - also available on the MVSEP site for free) used. 

Is that the same as this?

https://pypi.org/project/BS-RoFormer/