How drum apps score you

Updated 3 August 2026

Software can judge your drumming three ways: by listening to it, by reading a MIDI stream from an electronic kit, or by watching it on video. Each one trades something away — what it can sense, what it has to guess, and how precisely it can place a hit in time.

None of the three is the accurate one. A microphone can hear a snare it was never told to expect; a MIDI module knows which pad fired but needs that pad to exist; a camera can see your grip and not the milliseconds that separate on-time from late. So the question worth asking about any of them is not which one scores highest. It is what the thing is measuring, and how wide its window is.

Microphone and audio onset detection

A microphone-based app listens for a transient — a sudden spike in the audio signal — and treats it as a hit. Its biggest advantage is reach: it works with any acoustic kit, with no cables, sensors or hardware beyond a phone or a microphone.

The cost is inference. Nothing tells the software which drum was struck; it has to guess from the sound’s shape and frequency. A rimshot and a crash on the same beat can smear into one transient, or split into the wrong two. Timing resolution is bounded by how fast a transient can be located inside the audio itself, and a sixteenth-note hi-hat under a ride still ringing out makes that location fuzzier again.

MIDI from an e-drum module

An electronic kit’s module already knows which pad was hit and how hard — it has to, in order to trigger the right sound. Reading that MIDI stream leaves nothing to infer: no guessing which drum, no separating overlapping hits out of a waveform, no transient shape to misread. A Roland TD-17 and an Alesis Nitro Max send the same kind of message about it.

The trade-off comes upfront rather than ongoing: you need an e-drum kit, or an acoustic kit fitted with triggers and a module. iQuantize is built on this approach for exactly that reason. How hard you hit is a number the pad sent; when you hit is the hardware timestamp attached to that message as it reaches your device. Neither is reconstructed afterwards out of a sound. Which kits and modules are supported covers what that needs at your end.

Video and computer vision

A camera sees what the other two cannot: grip, posture, stick height, how the body moves into a stroke. That makes it genuinely useful for technique review in a way neither audio nor MIDI attempts to be.

It is weak exactly where drumming lives or dies: the millisecond-level timing that separates a groove that sits in the pocket from one that does not. Frame rate and motion blur put a floor under how precisely a strike can be placed in time, and that floor sits well above what feel actually requires.

What happens to your hit after it is detected

However the hit was sensed, three decisions come next, and you are the one who lives with them. Your hit gets compared to a grid. A window decides how far off still counts as on time. And something has to average all of it into the number at the top of the screen. Those three choices move your score further than the sensing method did.

A 100 ms window makes almost anyone look accurate — most human timing errors are smaller than that, so a wide-enough window simply stops registering them. And averaging every hit equally lets a busy part hide a struggling one: a hi-hat playing constant sixteenth notes generates far more hits than a sparse four-note kick pattern, so a kick that is consistently late barely moves the average at all. iQuantize instead averages each voice equally — the kick’s score counts the same as the hi-hat’s, regardless of how many notes either one has.

How to read any app’s score honestly

Before trusting a number, ask three questions. What is it actually measuring — audio, MIDI or video? How wide is the window behind “on time”? And what does it do when it cannot measure something cleanly: show a low score, or say so? Those are the three questions we answer about our own scoring, window sizes included.

See the MIDI approach in practice

iQuantize reads your kit’s MIDI directly and scores each voice on its own, so a busy hi-hat cannot cover a late kick.

App Store Download for Mac

Requires an electronic drum kit or a MIDI-capable drum module.

Common questions

Which method gives the most accurate timing?

MIDI from an e-drum module, because a pad firing is an event rather than something to be recognised in a recording. The message arrives with a hardware timestamp attached. Audio has to locate a transient inside a waveform, and video is limited by frame rate, so both carry more uncertainty around the millisecond level that timing scoring depends on.

Can a microphone-based app work with a normal acoustic kit?

Yes, and that is its main strength — no hardware beyond a phone or microphone. The trade-off is that it has to guess which drum made a sound rather than being told, which shows up most on simultaneous hits and in a noisy room.

Why does the hit window matter more than an app’s accuracy claim?

Because the window defines what counts as accurate in the first place. A 100 ms window will pass almost any human timing error, so two apps can both claim a high accuracy percentage while judging against completely different standards.

Does averaging every hit equally give a fair score?

Not necessarily. If a busy voice like hi-hat sixteenth notes is averaged in with a sparse one like a four-note kick pattern, the busy voice dominates the number and a consistently late kick can go almost unnoticed. Averaging each voice equally, rather than each hit, keeps a quiet part from hiding a real problem.

Can computer vision replace MIDI or audio for scoring?

Not for timing. Video is genuinely useful for things audio and MIDI cannot see — grip, posture, stick height — but frame rate and motion blur put a floor under its timing precision that sits above what feel actually needs.