Mid-Side Processing Explained: What It Does and What It Cannot

Mid-side processing has an odd reputation. Mastering engineers use it constantly and quietly. Meanwhile it became famous on the internet for something it was never good at — removing vocals — which is probably why most people's first encounter with M/S was a disappointing free tool that made a song sound like it was playing through a wall. Both things are true, and it is worth separating them.

Stereo is not two speakers. It is a field.

Play a record and close your eyes. The snare sounds like it is directly in front of you. A guitar sits off to the left, strings spread wide across the back, the singer squarely in the middle.

None of those positions physically exist. There are only two speakers. Everything between them is an illusion your brain constructs from the differences between what each ear receives.

A mixing engineer builds that illusion deliberately. Put a sound equally in both channels and you hear it centred. Put more of it in the left and it drifts left. This is why the vocal, kick and snare — the spine of most records — usually sit dead centre, while guitars, keys, backing vocals and reverb are spread outward to create width.

Rotating the axes

Left and right is not the only way to describe a stereo signal. Mid-side describes the same thing along a different pair of axes.

Mid is what the two channels have in common: (Left + Right) ÷ 2. Everything centred survives, everything panned partly cancels. Mid is the middle of the image.

Side is the difference between them: (Left − Right) ÷ 2. Anything identical in both channels subtracts to nothing, so centred sound disappears. What remains is width — the spread instruments and the reverb.

The elegant part is that this is fully reversible. Add mid and side back together and you get the left channel; subtract and you get the right. Nothing is lost in the conversion itself. That is what makes M/S so useful: you can convert, adjust, and convert back, with the only change being the one you intended.

Where mid-side genuinely shines

This is the part that gets overlooked, because M/S is mostly used quietly in mastering rather than as a headline feature.

Controlling low end without losing width. Bass energy spread across the stereo field causes real problems — it eats headroom, and on a club system or vinyl cut it can be genuinely destructive. In M/S you can roll off everything below roughly 120 Hz on the side channel only. The bass collapses to the centre where it belongs, the mix gets tighter and louder, and the width up top is completely untouched. You cannot do this with ordinary left-right EQ.

Widening without artificial effects. Nudging the side channel up a decibel or two makes a mix feel more open, and it sounds far more natural than a stereo-widener plugin, because you are raising material that genuinely exists rather than synthesising fake difference.

Treating a vocal without touching everything else. A little compression on the mid channel steadies a centred vocal while leaving the wide instruments alone. It is a surprisingly transparent way to control a lead.

Checking mono compatibility. Solo the side channel and listen. Anything loud there will weaken or vanish when your track is played in mono — which is what happens on phone speakers, most Bluetooth speakers, and many club systems. Engineers use this as a quick sanity check before signing off a master.

Why it became the standard vocal-removal trick

The reasoning is easy to follow. Vocals sit centred. The side channel removes centred content. So the side channel is your instrumental.

For a while that was the only browser-friendly option available, and it costs almost nothing to compute — which is exactly why so many free tools used it.

And why the results were always disappointing

Every flaw follows directly from the mechanism, which is why no amount of tweaking ever fixed it.

The rhythm section leaves with the singer. Kick, snare and bass are centred in almost every record ever made. Cancel the centre and you cancel them too. This is why those instrumentals always sounded thin and spineless — the low end was simply gone.

The stereo image collapses. You are keeping only the difference signal, so the result is no longer a stereo mix. It is a narrow, phasey, hollowed-out thing. That boxy "inside a tin" quality is not a bug in a particular tool; it is what the maths produces.

The vocal does not fully leave. Lead vocals are almost always given reverb or delay, and those effects are deliberately spread wide. The dry voice cancels; its reverb does not. So a ghost of the singer wobbles underneath the track — often the most irritating artefact of all, because it is unmistakably the thing you were trying to remove.

Mono files cannot work at all. If both channels are identical, the difference is silence. There is nothing to extract. The method does not degrade gracefully here — it fails completely.

It cannot give you drums. This is the fundamental ceiling. M/S knows only two categories: centred and not-centred. It has no concept of what a drum is, so it can never hand you drums as their own track. Real stem separation is simply out of reach.

What replaced it

The breakthrough was to stop asking where a sound sits and start asking what it is.

A modern separation engine is trained on a large body of music where the individual layers were known separately. Over many examples it learns the character of a voice, a snare, a bass — how each behaves over time, not merely where it was panned.

Because it recognises sounds rather than positions, it keeps the kick drum while removing the singer even though both are centred. It preserves the stereo image instead of destroying it. It handles mono recordings perfectly well. And it can separate drums and bass into genuinely distinct tracks, which the old approach could never do at any quality.

That is the approach behind the 7By.in Vocal Remover and Stem Splitter. Mid-side remains a fine mixing and mastering tool — it just turned out to be the wrong instrument for this particular job.

Mistakes people make with M/S

Because it feels powerful, mid-side invites overuse. A few patterns come up repeatedly.

Widening everything. Pushing the side channel hard makes a mix sound impressive in isolation and hollow in context. The centre is where your vocal, kick and snare live — the parts carrying the song. Weaken the centre and you have traded impact for novelty. If a mix sounds wide but somehow limp, this is usually why.

Forgetting mono playback. Phone speakers, most Bluetooth speakers and many venue systems collapse stereo to mono. Anything you added to the side channel partially cancels when that happens. A mix tuned for maximum width can lose noticeable energy the moment it plays somewhere real. Always check in mono before deciding you are finished.

Boosting side reverb without noticing. Reverb tails already sit mostly in the side channel, so raising side level raises reverb more than the instruments you were aiming at. The mix gets washier rather than wider, and it happens gradually enough that you may not spot it.

Treating it as a fix. M/S adjusts balance between centre and sides. It cannot repair a mix where the arrangement itself is crowded. If instruments are fighting, the answer is arrangement or EQ, not stereo manipulation.

Try hearing it for yourself

If you have any DAW, this takes five minutes and teaches more than reading about it.

Load a commercial track. Duplicate it, flip the polarity of one copy, and sum them to mono — you are listening to the side signal. The vocal will thin out dramatically, the drums will lose their punch, and the whole thing will sound small and strange. That is precisely the sound of every old free vocal remover, and once you have heard the mechanism directly, the limitation stops being abstract.

Then run the same track through a modern separation tool and compare. The difference is not subtle.

Common questions

Is mid-side processing outdated?

Not at all — it is used daily in mixing and mastering, particularly for low-end control and mono checks. It is outdated only as a method of separating instruments.

Does M/S conversion damage audio?

No. The conversion is mathematically reversible, so converting to M/S and straight back leaves the audio identical. Damage comes from what you do in between.

Why do some free vocal removers still use it?

Because it is nearly free to compute. That is the whole reason, and it is also why those results still sound the way they always did.

Can M/S give me just the drums?

No. It has no concept of instruments, only of position. Use the Stem Splitter for that.