Under the hood

The KimiCue
Algorithm

Most auto-cue tools find where your track starts. KimiCue finds where your track drops.

That's a fundamentally different problem - and it's the one that actually matters when you're standing behind the decks.

Multi-signal pipelineGenre-tuned profilesProprietary scoring formulaConfidence rated

Multi-signal detection

Amateur tools look at one thing.
KimiCue looks at six.

Amateur drop detection looks at one thing - usually loudness. That's why it gets it wrong.

KimiCue runs a multi-signal analysis pipeline across several layers of the audio simultaneously, combining them through a proprietary scoring formula tuned across real DJ tracks in real genres.

No single signal makes the decision. Every candidate drop point in the track is scored across all signals at once, weighted against each other, and ranked. Only candidates that pass every filter make it through.

Simple loudness detection

1 signal

Finds loud moments. Gets it wrong on breakdowns, intros, and anything with compression.

KimiCue scoring pipeline

Energy surge
88
Bass return
94
Spectral transition
78
Onset density
72
Phrase alignment
82
Sustained energy
90

All six signals combined into one score per candidate beat. Highest score that passes all filters wins.

What goes into the score

Six signals. Every candidate beat.

01

Energy surge

Overall loudness shift at the drop

Real drops hit harder than the breakdown before them. KimiCue measures exactly how hard - not just whether the track gets louder, but by how much and how fast.

Signal weight88/100
🔊
02

Bass return

Low-frequency layer analysed in isolation

A drop almost always brings the kick and bassline back in after a breakdown. The low-frequency layer is isolated from the rest of the track and analysed separately. Bass return is one of the strongest signals in the formula.

Signal weight94/100
03

Spectral transition

Full frequency character shift

A drop doesn't just get louder - it sounds fundamentally different to what came before it. The full frequency character of the audio is analysed frame by frame and the transition is measured across the entire spectrum.

Signal weight78/100
04

Onset density

Multiple elements arriving at once

Drops are defined by multiple sonic layers hitting simultaneously - kick, bass, synth, percussion all arriving together. KimiCue detects that clustering of new sound events.

Signal weight72/100
05

Phrase alignment

Musical boundary check

Music is written in phrases. Drops land on phrase boundaries - bars of 16, 32, 64 or 128 beats. The algorithm scores candidates higher when they fall where a real drop structurally should.

Signal weight82/100
06

Sustained energy

The drop has to hold

A moment that spikes and immediately falls is a false positive. KimiCue requires sustained energy after the candidate point before it will call it a real drop. This single filter eliminates the majority of false positives that simpler tools fall for.

Signal weight90/100

Genre-tuned profiles

Same signals. Different formula.

The same audio signals mean different things in different genres. KimiCue uses individually tuned detection profiles for every supported genre - each one specifying different signal weights, different search windows, different phrase expectations, and different confidence thresholds.

These profiles were developed through analysis of real tracks across each genre, not estimated from theory. When you select a genre mode you're not filtering a list - you're switching the entire scoring formula to one calibrated specifically for the music you're working with.

🌊

Trance

Drop timing

Late - often after 90s+

Key signal

Long sustained plateau

Build lasts a long time. Drop defined by atmosphere and sustained energy, not an aggressive bass hit.

Drum & Bass

Drop timing

Early - ~30 seconds in

Key signal

Sharp aggressive bass hit

Fast-moving structure. Short intro, hard and immediate drop. Bass return signal is dominant.

🎵

Melodic / Progressive

Drop timing

Very late - often 3+ minutes

Key signal

Atmospheric energy shift

Build can last three minutes. Drop defined more by atmosphere and spectral change than by bass.

Techno

Drop timing

Gradual - 4–8 minutes in

Key signal

Slow incremental energy build

Long drawn-out structure. No single moment of drop - energy rises slowly over many minutes.

Auto Detect works from your track tags

Auto Detect reads the genre tag embedded in your track metadata and selects the correct profile automatically - so for a well-tagged library you don't need to think about it at all. Make sure your tracks are tagged correctly in Rekordbox before running for best results.

Studio-grade audio analysis

Built on professional signal processing.

KimiCue's signal processing is built on the same audio analysis foundations used in professional music research and studio tooling worldwide.

We work at a 22kHz sample rate with a 2048-point FFT window - giving us frequency resolution fine enough to isolate bass, mids, and highs as separate signals, not just overall loudness.

Beat tracking is performed against the full onset envelope of the track, not a simple BPM estimate - so the algorithm snaps drop candidates to real beat positions rather than theoretical grid lines. When KimiCue places a cue, it lands exactly on the beat, every time.

Professional audio analysis stackResearch-grade

22kHz sample rate

Full audio fidelity for analysis

2048-point FFT

High frequency resolution - bass isolated from mids and highs

Beat-snapped cues

Cues land on real beat positions, not grid estimates

Onset envelope tracking

New sound events detected from spectral flux, not amplitude

Per-genre confidence thresholds

Pass/fail criteria tuned for each music style

Honest about what it does

We'd rather be straight with you.

Best on structured electronic music

Tracks without a defined drop - ambient, live recordings, heavily experimental music - may return a low confidence score and get skipped. That's by design. A skipped track is better than a wrong cue in your set.

📊

Confidence score tells you when to check

After each run, KimiCue shows a confidence score for every track. High confidence is ready to use. Lower scores are worth a quick check before you rely on them in a live set.

🔁

See it before you commit - Dry Run

Run a full analysis in dry run mode first. It shows you exactly what the algorithm detected, track by track, without writing a single cue point. The best way to see the scoring on your own music before anything changes.

💡

Why KimiCue

How it compares to Rekordbox auto-cue and other tools.

Genre Mode & Cue Template

Pick the right settings for your music.

📖

Full guide

Settings, Serato, license - everything in one page.

Select Language