Coming first to macOS. VST3 and AAX, on Mac and Windows, to follow.
Why layering goes wrong
Lining up the first hit is not enough. Two drums almost
never ring at the same rate, so a sample that starts in step drifts out of it within a
few cycles. What you layered for punch quietly takes it away.
How it works
Two waves. One push.
Lined up. Still fighting.
The transients start together, but this sample is upside down against the drum.
Add the two and the hit gets smaller.
Flipped.
Now the first cycles push together. The sample rings a little fast, so by the
tail it has drifted back out of step.
Retuned by −90 cents.
Played slightly slower, like tape, it rings at the drum's rate and stays in step
to the end of the hit.
Live drumSampleTogether
Polarity
Inverted
Tuning
0 cents
Coherence
0.00
An illustration, measured as it draws. Coherence 1.0 means unrelated, 2.0
perfectly in step. On a real snare pair from the test set, Drum Match took an
untouched sample from 0.758 to 1.759.
Listen
Hear it.
A live snare with a sample layered under it. Same hit, same sample.
The only difference is the match.
Tuning
−135 cents
Coherence
1.759
Level
Matched
Both versions play at equal loudness, the way the blind test
was run, so you hear the match rather than more level. Turn it off to hear the natural
difference.
Example 9 from the test set: a close-miked Pearl snare. The matched sample
is Drum Match's own output, not a hand match. Headphones help.
What it changes
Three things. Nothing else.
1
Speed
Faster or slower, like tape, so the sample rings at the drum's rate. Not a pitch
shift: the length has to change for the waves to line up.
±1200cents searcheda full octave each way
2
Polarity
When a sample is the mirror of the drum, flipping it turns cancellation into
reinforcement.
+7.7 dB from the flip alone, on one real pair.
3
The low end, separately
Body and weight can want opposite things. Drum Match splits just under the drum's
own pitch and decides each half on its own. The split adds back to exactly the
original.
0.504 → 1.496 low-band coherence, worst pair.
Never touched
Level stays at unity. Timing stays yours beyond aligning the transient. The live
drum is never processed at all.
Knows when to say no
Some tuning always scores best. That does not make it good. If the curve has no
clear peak, Drum Match reports No match and will not
arm playback.
Four drums, four slots
Snare
Kick
Rack tom
Floor tom
Close-miked. Up to four samples per drum, each matched on its own.
Tested by ear
13 of 13.13 of 13.
In a blind test on thirteen unselected snare pairs, the matched
sample was judged better than the untouched one every time.
The story
Why I built it.
I've loved music since I was a kid, and these days I mix for a living as an audio
engineer. Layering a sample under a live snare should add weight, but half the time it
took some away, and the only fix was nudging tuning and polarity by ear, hit after hit,
until the two stopped fighting.
Drum Match does that part. It measures what I was listening for, and it was tested
the way I would have tested it: blind, by ear, against my own matches.
JoelPolar Bear Audio
Get on the list.
One email when Drum Match is ready. No account, no card.
Your address is used to tell you when Drum
Match is ready, and to invite you to test early builds only if you ask. Unsubscribe is
in every email.
It is for layering. Drum Match works on the sample you put under a live drum, so the
two add up instead of fighting. The recorded drum is never touched.
Is it pitch-shifting the sample?
No. It resamples, like changing tape speed, so length and pitch move together. The
length is exactly what has to change for the waves to line up. Every hand-made
reference match we studied turned out to be plain resampling.
Will it change my levels?
No. Gain stays at unity. In every reference match an engineer made by hand, the
level was never changed either.
What happens when nothing fits?
It says so. If the reinforcement curve drifts instead of peaking, the best number is
only less cancellation, not a match. Drum Match reports No match and leaves that slot
unarmed.
Does the analysis run in real time?
It runs once per pairing, off the audio thread. During playback the plugin only plays
a sample that has already been matched.
Formats and systems?
First release: Audio Unit on macOS, which covers Logic, Ableton Live, Reaper, Studio
One and most other Mac DAWs. VST3 (Cubase, Nuendo) and AAX (Pro Tools) follow, then
Windows. Tell us your DAW when you sign up and it helps us decide the order.
Keep your cool.
Drum Match is coming to the Mac first. One email when it is ready.
