Drum Apps That Listen to Your Playing: What They Actually Measure
Tools · Listening apps · Updated 2026-08-16
A drum app that listens uses your phone's microphone to detect each stroke, timestamps it, and compares that timestamp to where the metronome grid said the note belonged. The output is a signed offset in milliseconds for every hit: negative for early, positive for late. That is the entire category. The scores, the colors, the histograms and the level ladders are all just ways of presenting a list of offsets.
This is the buyer's guide and troubleshooting manual for that category. It covers what these apps can and cannot measure, why an accuracy percentage means nothing without the window it was measured against, and the setup mistakes that make microphone scoring lie to you: speaker bleed, Bluetooth latency, and thresholds that quietly swallow the second stroke of every double. For the underlying numbers, start with what a millisecond of timing error actually is.
What it means for an app to listen
Every app here does the same three things. It runs a metronome, so it knows where each note should land. It watches the microphone for onsets, the sharp rise in amplitude a stick makes on impact. Then it subtracts: detected stroke time minus expected grid time. The answer is a signed offset in milliseconds, negative for early and positive for late.
Everything after that is arithmetic. A stroke inside the app's on-time window counts as on time. Outside it and early, it is rushed. Outside it and late, dragged. A grid slot with no stroke near it is a miss, and a stroke with no grid slot near it is an extra. Those five labels are the raw material behind every score, color and graph you will see. When a listening app behaves strangely, it is almost always because the onset detector is seeing something other than your sticks, which makes the fix physical rather than musical.
This is a category, not a single product
Search results make this look like a one or two app market. There are at least ten. Ticking, Score Drummer, Drummer ITP, StickPulse, EarDrum, Rudimentalist, DrumPulse, Drumstik and Drumr all work here. Several advertise per-hit early and late feedback, and several process audio on the phone. Any app claiming to be the only one that listens to drumming, or the first, is overstating. Drum Coach₊, which we make, is one of many. What actually separates products is how much of the measurement they show you.
What these apps measure, capability by capability
A product-by-product comparison goes stale within months, so compare by capability instead. The last column is the part vendors leave out.
| Capability | What it measures | How it is usually presented | Honest limit |
|---|---|---|---|
| Onset detection | Whether a stroke happened at all, and how many | A hit counter, or notes lighting up as you play | Says nothing about timing, hand or volume. If this is wrong, every number built on it is wrong too. |
| Per-hit timing offset | Signed distance in milliseconds between each stroke and its grid position | A plus or minus figure per note, or color coded note heads | A single offset is noise. Only the pattern across a full rep carries information. |
| Rush and drag split | Whether the errors sit early or late, and by how much on average | Two counts, or a mean signed offset with a direction | A near zero average can hide large errors in both directions that cancel each other out. |
| Spread and consistency | How widely the offsets scatter around the grid | A histogram, a scatter plot, or a standard deviation in ms | Needs enough strokes to be stable. A four bar rep is a small sample. |
| Hand to hand evenness | Whether alternating strokes divide the beat equally, right compared with left | Two distributions side by side, or a single evenness figure | One microphone infers the hand from the printed sticking. It does not hear which hand played. |
| Accent and dynamic discrimination | Relative stroke loudness checked against the written accent pattern | Velocity bars, or accents marked as hit or missed | Relative only. Room, mic distance and pad all change the absolute level. |
| Missed and extra strokes | Grid slots with no stroke, and strokes with no grid slot | Flagged in the stroke timeline | Cannot distinguish a note you failed to play from a note the microphone failed to hear. |
| Overall accuracy score | Share of strokes falling inside the app's on-time window | One percentage | Meaningless unless the window in milliseconds and the tempo are printed beside it. |
| Tempo progression | Whether you have earned the next tempo | A ladder that unlocks, or a timer that advances regardless | A timer based ladder measures patience, not accuracy. |
Two rows deserve expansion. Hand attribution from a microphone is inference, not detection. A phone mic hears one blended signal and cannot tell a right hand from a left. Apps that report per-hand data know the rudiment's sticking and map each stroke to its position in the pattern, so in a single paradiddle written R L R R, stroke one is a right and stroke two a left. Velocity is relative, not absolute: a mic can rank strokes loud to soft within a phrase and check them against an accent pattern, but it cannot give a level that transfers between rooms.
Why an accuracy percentage means nothing on its own
If an app says you played 92 percent accurate, the first question is: against what window? Ninety-two percent inside 50 ms and 92 percent inside 15 ms describe two different players. The window is the claim; the percentage is a consequence of it. A tool worth using states its window and ideally lets you tighten it as you improve. Drum Coach₊ defaults to a 25 ms on-time window, but even that only means something next to the gap between consecutive strokes.
That gap shrinks as tempo rises while a fixed window stays the same size, so the window swallows a larger share of the space between notes the faster you go. Sixteenth notes are four strokes per beat, so the spacing is 60000 divided by BPM times four, that is 60000 / (BPM x 4).
| Tempo (BPM) | 16th note spacing (ms) | 25 ms window as share of the gap | Strokes per minute |
|---|---|---|---|
| 60 | 250 | 10.0% | 240 |
| 100 | 150 | 16.7% | 400 |
| 120 | 125 | 20.0% | 480 |
| 140 | 107.1 | 23.3% | 560 |
| 160 | 93.75 | 26.7% | 640 |
| 180 | 83.3 | 30.0% | 720 |
| 200 | 75 | 33.3% | 800 |
Read the third column as slack. At 60 BPM a 25 ms window covers a tenth of the space between two sixteenths, so passing it is a real claim about steadiness. At 200 BPM it covers a third of that space and the same percentage is a much softer claim. That is why a score at one tempo is not comparable to a score at another, and why a tempo ladder feels harder near the top.
Why microphone scoring goes wrong, and how to fix it
This is where most people abandon listening apps, usually for a reason that has nothing to do with their playing. A microphone is a blunt instrument: it hears your sticks, your click, your room and the surface the phone rests on, and it cannot tell them apart. Work through this table before you believe any score.
| Symptom | What is actually happening | Fix |
|---|---|---|
| Score collapses to near zero while the playing feels solid | The phone speaker is playing the click and the microphone is hearing it. Each click is counted as a stroke, so the app sees roughly double the strokes and flags half of them as extra. | Use headphones, wired if you have them. If you must use the speaker, drop the click volume, move the phone off the same surface as the pad, and choose a click sound with a short, bright attack. |
| Every stroke reads late by roughly the same amount | Output latency. The app timestamps the click when it generates it, but you hear it milliseconds later and play to what you heard. Bluetooth is the usual culprit. | Run the app's latency calibration, then run it again after connecting or disconnecting headphones. An app that stores an offset per output route and re-runs on route change handles this for you. |
| Double stroke rolls score near zero while singles score fine | The second stroke of each double is quieter than the first and falls below the onset detection threshold. The app logs a miss rather than a late note. | Move the phone closer, play on a harder or less damped surface, lower the sensitivity threshold if the app exposes one, and deliberately overplay a few second strokes to confirm the detector can see them at all. |
| Extra strokes appear that you never played | Room noise crossing the threshold: a fan, air conditioning, a footstep, a door, music through a wall, or a second drummer in the room. | Move to a quieter room and put the phone closer to the pad so real strokes sit further above the noise floor. Stop scoring when someone else starts playing. |
| Changing pads changed the scores overnight | Pad surface changes the attack the mic sees. Soft, heavily damped gum rubber gives a low dull onset; a harder surface gives a sharp transient that is far easier to detect. | Score on the harder or brighter surface and keep the same pad every session. Save the quiet pad for late night volume, and expect more missed strokes when you use it. |
| An acoustic snare scores worse than a practice pad | Decay, buzz and sympathetic ring smear the attack, so onset timing gets less certain and rattle can trip the threshold on its own. | Use a pad or a muffled head for scoring sessions. Play the snare for music, not for measurement. |
| Scores swing between sessions with no change in your playing | The measurement setup moved, not the playing. Mic distance, angle, table surface and phone case all change how loud a stroke looks to the detector. | Fix one phone position and one pad and keep them. Mark the position if you need to. Only compare sessions recorded in the same setup. |
| Counts go wrong only at high tempo | Onset detectors enforce a minimum gap between hits to avoid double triggering, and at fast tempos consecutive soft strokes inside a roll can fall inside that gap. | Drop the tempo until the stroke count is right, then climb again. If the count is wrong, every timing number above it is fiction. |
The five minute setup that makes the numbers trustworthy
- Place the phone, do not bury it: Stand the phone about an arm's length from the pad, mic end facing the pad, nothing over the bottom edge. Do not lay it on the pad itself: it picks up body vibration through the surface and smears every attack.
- Get the click out of the room: Headphones are the biggest reliability upgrade here, and wired ones add the least delay. Speaker playback puts the click into the same air the mic is listening to, and speaker bleed is the top cause of nonsense scores.
- Calibrate latency on every route change: Wireless audio delays the click on its way to your ears, which makes an accurate player look like a consistent dragger. Recalibrate whenever you switch between speaker, wired and wireless output.
- Prove the detector sees singles: Play twenty slow singles at 60 BPM, where one beat is a full 1000 ms, and check the app counted twenty. If the count is wrong, nothing built on it is worth reading.
- Prove the detector sees doubles: Now play a slow double stroke roll as R R L L and count again. The second stroke of each double is quieter and is the first thing a threshold discards. If half of them vanish, that is a microphone problem, not a technique problem.
- Fix the window, then read the score: Set the on-time window once and leave it. Changing it changes the score, so a moving window makes your history meaningless. Tighten it only when you can hold a target at the current setting.
Where a microphone genuinely loses to a sensor
Be honest about the input. A microphone is free, universal and works with the pad you already own. It also has limits a trigger pad does not: sensors give a separate signal per pad, a genuine velocity value per stroke, and near immunity to room noise and click bleed, which is why platforms built around MIDI hardware can measure things a phone cannot. If you already own an e-kit or a trigger pad, prefer an app that accepts it. Otherwise microphone scoring is genuinely useful once set up as above, provided you read it as a trend across sessions rather than a verdict on one rep.
The buyer's checklist
Eight questions separate a measuring instrument from a scoring toy. Most are answerable from a store listing; the rest show up in the first five minutes.
- Does it show a distribution or a single number? One percentage cannot separate a player who is consistently 30 ms early from one scattered randomly by the same amount, and those two need different practice.
- Does it separate rush from drag? An unsigned error figure throws away direction, the most actionable part of the measurement, and rushing and dragging have different fixes.
- Does it measure each hand? Most rudimental problems live in one hand. An app that reports right against left tells you which hand to isolate, the fastest route to even double strokes.
- Does it state its on-time window, and can you change it? A percentage without a window is decoration. If the listing never says what counts as on time, assume the window is generous.
- Does tempo advance on measured accuracy or on a timer? A ladder that climbs because a countdown ended is measuring patience. A gated ladder is measuring you, and that is the difference between a tempo chart you earn and one you scroll.
- Does it handle your headphones? Ask whether it calibrates output latency and recalibrates when the route changes. Without that, wireless headphones will make you look like a dragger forever.
- Is processing on-device? Prefer apps that analyze locally and state plainly that nothing is recorded, saved or uploaded. Several apps in this category do it, so there is no reason to accept less.
- One-time purchase or subscription, and what sits behind the paywall? Both models exist here. The question is not which is cheaper but whether what you are buying keeps changing.
That last point deserves a straight answer. A one-time purchase suits a tool that is finished: a metronome, a rudiment reference, a scoring engine. None of those need a monthly content pipeline, and paying every month for a static list of 40 rudiments is poor value. A subscription suits a product with an ongoing production cost, such as a lesson platform licensing songs and filming new courses. Work out which side of that line an app sits on, then ask what happens to your practice history if you stop paying.
Measure the strokes, do not guess at them. Drum Coach₊ by Panda Taps, coming soon for iPhone, is one of several apps that score drumming through the microphone. It classifies every stroke as on-time, rushed, dragged, missed or extra against a stated default window of 25 ms, shows the result as a per-stroke timing histogram rather than one opaque number, compares your right hand against your left on double strokes, and moves you up a 12 rung BPM ladder only after two reps at 90 percent or better at the same tempo, never on a timer. First-run latency calibration stores an offset per output route and re-runs when the route changes, so wireless headphones behave. All 40 PAS rudiments come with notation, sticking, accents and audio demos, free. Audio is analyzed on the device and never recorded, saved or uploaded, with no account, no ads and no analytics. Learn more about Drum Coach₊.
Frequently asked questions
Is there an app that listens to your drumming and tells you if you are rushing?
Yes, and more than one. Apps including Ticking, Score Drummer, Drummer ITP, StickPulse, EarDrum, Rudimentalist and Drum Coach₊ work in this category, and several report early or late per hit. What differs is whether you see a single score or the full spread of offsets underneath it.
Why does my double stroke roll score so badly in a drumming app?
Usually because the second, quieter stroke of each double falls below the microphone detection threshold, so the app records it as missed rather than late. Test it by playing slow doubles and counting the registered strokes. If half are missing, move the phone closer, use a harder pad surface, or lower the sensitivity threshold.
Do these apps work with Bluetooth headphones?
Only if the app calibrates for output latency. Wireless audio adds delay between the app generating the click and you hearing it, which makes every stroke look late by roughly the same amount. Apps that measure and store an offset for each output route handle it. Apps that do not will read accurate playing as constant dragging.
Can a phone microphone tell my right hand from my left?
Not on its own. A single microphone hears one blended signal and cannot identify which hand played a stroke. Apps that report per-hand data infer it from the rudiment's known sticking, mapping each detected stroke to its position in the pattern, which stays accurate only while you play the printed sticking and every stroke is detected.