What happens in the separation
A finished track is a single audio file — the individual instruments are no longer separate in it, they are summed together. Separation does not restore them, it estimates them. A model decides for every moment which share of the sound belongs where and builds four new parts out of that.
Everything else follows from that. Where an instrument stands clearly on its own, the estimate is very good. Where two sounds overlap — a reverberant vocal over a pad, a distorted bassline that half sounds like a synth — the model has to guess, and then residue remains. That is not a malfunction, it is how the thing is built.
How to hear a separation
Four signals give it away almost every time. Knowing them means deciding before the gig which tracks are suitable — not in the middle of a transition.
| Signal | When it shows up | What you can do |
|---|---|---|
| Residue in the vocal | On dense productions with a lot of reverb | Use it briefly instead of letting it run. Nobody hears it in a transition; across a whole verse they do. |
| Thin bass | When bass and kick sit in the same range | Instead of removing the bass entirely, pull it back with the EQ — that often sounds more natural than the separation. |
| Smeared highs | On old or heavily compressed recordings | Material like that is not suitable. You notice it in ten seconds when preparing and too late in a set. |
| Dropouts | When two decks are separated at the same time | Separation costs processing power. On older machines use one deck at a time, and test under full load at home first. |
What the programs ask for it
The function is now widely available, but not everywhere on the same terms: Traktor introduced separation with Pro 4, and in rekordbox the feature set depends on the plan you are on — which tier contains what is listed in the manufacturer's pricing overview and changes faster than any guide. Before buying, look at the current page, not at a video from two years ago.
More important than who has it is where it runs. The separation happens on your machine and costs processing power — on older hardware that becomes audible as dropouts when two decks are separated at once. If you want to work with it in a club, test it at home under full load and not in a set.
Using stems in six steps
The order in which this fits into a set without accidents. The first three steps happen at home.
Try it at home on twenty tracks
Not on one. Only across a handful do you notice which kind of production separates well and which does not — and that is exactly the knowledge you will miss in a set.
Tag the tracks where it works
A tag in the library, a word in the comment field. In a club you want to know this, not test it.
Check the load with two decks
Both decks running, both separated, effects on. If it crackles here it will crackle in the club too — only with an audience.
Start with one element
Take out the bass, nothing else. One element, one effect — switch three parts at once and you can no longer hear which one made the difference.
Put it on a phrase boundary
A part that disappears mid-section sounds like a mistake. On the downbeat the same move sounds deliberate.
Bring it back
Taking something out is half the move. When it comes back decides whether the room experiences tension or a hole.
The five mistakes you hear immediately
Removing the vocal in every transition
What happens in every transition is no longer an effect, it is a noise. After twenty minutes the room stops hearing it as an idea and starts hearing it as a habit.
Using stems instead of the EQ
For a bass swap the EQ is faster, more reliable and sounds better. Separation is meant for what the EQ cannot do — individual elements rather than frequency ranges.
Finding out live whether a track is suitable
Ten seconds of listening at home saves you the moment in a set where a vocal suddenly sounds tinny. That belongs to preparation, like setting cue points.
Separating two decks on weak hardware
The processing load rises with every separated deck. A dropout in the middle of a transition is the most expensive way to learn that.
Treating separation as a substitute for selection
A track that does not fit does not fit without its bassline either. Stems change how something sounds, not whether it belongs to this night.
The three uses that stick
A lot of what stems make possible goes unused after a month. These four things do not.
The switch underneath a vocal
You keep the vocal of the playing track and swap everything underneath it. That is the one use for which there was no solution before.
The bass you do not want twice
Instead of turning the EQ you remove the outgoing track's bass part entirely. Cleaner than any cut — as long as the material allows it.
The acapella that does not exist
For most tracks there is no official acapella. For four or eight bars over another instrumental the separation is entirely enough.
Tidying up in an emergency
When two tracks unexpectedly fight each other, removing one part is faster than any correction — as a rescue, not as a plan.
One judgement this guide makes deliberately: stems do not replace craft. A transition that does not work without separation does not work with it either — it just sounds wrong differently. The function is strongest in the hands of someone who already mixes cleanly without it, and most conspicuous in the hands of someone using it as a shortcut. That is not a moral, it is an observation you can test any night.
Frequently asked questions
What are stems in DJing?
The real-time splitting of a finished track into individual parts — usually four: vocal, drums, bass and the melodic remainder. Each part can be muted or removed on its own. Serato documents the controls for this in its manual; the other major programs now have it too.
How good is the quality really?
Much better than most people expect, and never perfect. On cleanly produced tracks with little reverb barely anything is left over; on dense productions, old recordings and heavily compressed material you hear residue. For a few bars in a transition it is almost always enough, for a whole verse over another instrumental often not.
Do I need new hardware for this?
Not necessarily, but you need processing power. The separation runs on your machine and the load rises with every separated deck. On older machines one deck often works fine and two do not. Test that at home under full load, not in the club.
Do stems replace the EQ?
No, they do something different. The EQ works with frequency ranges, separation with elements — for a bass swap those overlap, everywhere else they do not. For the swap the EQ is faster and more reliable. For switching underneath a held vocal there is no EQ move at all.
Which software has stems?
All the major ones by now. Traktor introduced separation with Pro 4, Serato documents the function in its manual, and in rekordbox it depends on your plan — the manufacturer's pricing overview says which tier contains what. This changes faster than any guide, so: the current manufacturer page beats a video.
Is that cheating?
The same question as with the sync button, and the same answer: the room does not hear how a transition came about. What gets noticed is not the function but the excess of it — a set where something is missing from every track sounds like a tool rather than like music.
Start learning
Mix first, separate later
On DJLearn you first learn the transitions that carry without stems — and after that, separation is a tool rather than a crutch. Free, and no account needed to try it.
Start for free