That’s going to be interesting.
I’d think that Frames Per Second is an important consideration, so splitting two layers, three ways each is going to mean 6 jumpy waveforms. One way around that might be to either disable one layer while stem play is active or if there are 4 stems (vocals, bass, piano, drums for example) then put two stems on layer A, two stems on layer b
I see that this thread was necrobumped. :-p
NI announced in 2018 there will be no continued development with stems.
Beatport announced in June that stems will be removed from their platform.
Something like neural mix would be the closest option.
Virtual DJ also has it. And it can be implemented on the Denon Prime 4 pads in controller mode using virtual DJ. It would be nice though to have stem separation on the Prime 4 in standalone mode.
How much latency are you willing to tolerate for something not many would use much after getting excited about it and trying it out?
Stems in real time is a game changer as it expands the art of DJing on so many levels adding to the creativity of a mix as there are many ways they can be implemented. Just think about it you can actually remix songs on the fly in so many different ways every time you play a song so that it doesn’t ever have to sound the same again.Adding or taking away different elements of another song added to it.To make the statement that very few people would use it is just your opinion. I come back from the time when we beat mixed with records. We only dreamed of remixing our own songs to the level that DJs are able to do it today.So as the technology in the DJ world improves I embrace every little bit of it and don’t take it for granted.
Just so you know the technique for doing this existed since to 80’s probably even before.
It take a 20 band eq and selective phase reversing. Algoriddim and vdj took advantage of this old Sokol technique.
With that said I can confidently tell you the way they did it is garbage. Everybody who tried it including me thinks it not as useful level as everybody wants it to be.
We also know that the processing gives unacceptable latency for live performances on real dj gear.
Maybe on your computers, where it takes a round trip of approx.12 ms at minimum, you find this acceptable but not for standalone.
I totally disagree. The technology is improving already. Virtual DJ now has the ability to control the amount of bleed coming through and as time goes on it will be perfected just as beat synk is much better… That being said how many times did you want a different mix of a song and you knew you would love the song only if you had a different mix of it… Real-time stems gives you the ability to remix your songs as never before. I remember when automatic beat synk first came out The DJ world scoffed at it saying it would destroy the art of DJing. Today we realize that’s not the case. It actually frees you to add creativity to your mixing. So don’t be so quick to dismiss this technology. You just have to use your imagination to figure out all the many ways it can be be implemented every day in your mixing.
Let’s put it to the test shall we?
Post a remix using algoriddim or vdj stems and let us backwards folk judge it’s quality.
There’s not many on the internet who praises it btw.
Unless it real stems, all you’ll get is fake sound that lost tonal character and sounds watery.
If you want fake sound try to go back to cassettes…haha. Again they made the same statement about automatic beat synk that it wouldn’t be used and it would be a waste of time. Now it’s standard on all DJ gear. And I’ll do you one better. Why would I want to limit your creativity. Go ahead try it yourself if you have Denon gear put it in controller mode with virtual DJ and start practicing, remember practice makes perfect. If you just use your imagination and give it enough time you will find many ways to implement it. But if you choose to shut your mind off to the possibilities then sure it’s not going to be any of use to you because you’ve chosen that for yourself. As a DJ why limit yourself. You deserve better than that.
It’s really bad sounding - it sounds like YouTube rips when VdJ separates the vocals and music etc - ok it’s clever that the algorithm can do any sort of half-arsed, half-baked attempt at splitting one sound type from another, but it’s not good quality results
Real stems operate exactly like the 8 track tapes of old except in the future you get 4 fake ones that don’t sound good 
I don’t think it sounds fake at all To me it just sounds like a filter is applied to it. Plus when you use the bleed function properly it makes for a cleaner sound. And if you have a little patience the technology will definitely get better it’s only a matter of time. Maybe then you might give it a chance.
The technology has long been here. Even our best military intelligence officers will tell you that to get the cleanest audio it has to be done post with resources beyond what’s available in the civil realm.
You will never get it clean.
The eqing will always take away the tone and the reverse phasing will always make the isolated sound filtered and phasey.
I can do a better job of it on my daw cause that’s how I was taught to do it in school. Do I want to do? No cause it’s not worth it knowing what the end result will be.
It’s good not everyone feels like you do if that were the case we wouldn’t have any of the DJ gear that we do have today.
Everything you see coming out as new features in the digital realm was already done in the analog form a very long time ago. Except the digital realm does it with unacceptable latency.
Unless these tracks are pre-prepared before hand with a lot of preprocessing nobody will get the illusion of real-time results with the clarity one desires.
Hope you understand that what you are asking for is a bigger undertaking than anyone wants. You need real stems period. Cause with real stems your software does not have to real-time process right? (Cause it’s already there) So now it has to dedicate resources to separate frequencies and isolate on multiple fronts this is not acceptable for real-time hi hq processing.
It’s only going to be a matter of time and all that will be implemented. It’s improving every year. If you don’t want to see that that’s your choice. There is plenty of demand for it out there Digital DJ Tips on YouTube is made up of a bunch of DJs from the UK who have many years of experience they’ve talked very positively about stems and how they will be implemented in DJ software in the future… So don’t believe me but at least believe the professionals.
I appreciate your faith but nobody can break the laws of physics as of now. And I don’t think anyone will break it in the future.
Well when they do my friend I’ll be the first to tell you I told you so…haha
We had a clean way of doing it and that was NI’s STEMS.
Nobody really wanted it as although they have away the technology to use, it needed VDJ, Pionee DJ and Serato to incorporate it to make it work.
It never happened, even as a free technology with free STEM making tools and with a £0.50 premium on tracks.
The major pools could have supported it by incorporating clean, dirty, instrumental and acapella versions to each track and you switch around the mix you want.
It never happened.
I had a Kontrol S8 and loved it as it was the same workflow as my old Novation Twitch. It was amazing to use but only for certain types of music and that is electronic. That killed it right there.
Using AI to separate the tracks audio into stems is okay but very hit and miss. I did an old Motown track and the results was an amazingly clean acapella. Perfect. I then tried a more modern track and it sounded like a dodgy YouTube rip.
The pressures of the analysis is fine as it will do it all when the track is scanned in Engine Prime first. The limitation would be playing a track that hasn’t been scanned in EP first like streaming platforms. I can’t see the unit doing the analysis without taking twice as long and ramping up its small CPU a fair bit.
Honestly, I can’t see it being too useful for my own situation but I can see where others I’ll use it. Grabbing an acapella to put an effect on it is fine but a raw stem will leave an ‘unpolished’ sound if we’re not too careful.
I can tell you 100% that politics played a major factor.
For example, I made a hot track and uploaded as a proper stem file. What’s stopping someone from using one of my stems and making a hit of it but don’t want to pay me?
You can see immediately where why stems never took of. 
To bypass this other software are separating frequencies and keeping phase distortions so it has no resemblance to the original.
Food for thought