How are these two tracks connected? Motif looks for shared recorded audio (a sample,
a re-release) and for musical resemblance (a cover or interpolation). It runs entirely in this browser,
using a C engine compiled to WebAssembly, so your files are never uploaded.
Where they connect
Each track's waveform, spectrogram (pitch up, time across) and chroma (the 12 pitch classes, C at the bottom).
Linked bands show the passages Motif matched. Click anywhere on a track to play from there; click a band to hear
the matched passage in both tracks together.
Shared recorded audioStrongest musical (chroma) match
Matched passages
Play a matched passage in both tracks at the same time, or one after the other. Change the start
and end to trim it, then save that part of either track as a WAV file.
Why this result
The detectors use fixed thresholds (dashed red lines). Short, pitch-shifted or heavily processed
samples and very dense mixes can fall below them, so a "no connection" result is not proof of absence. These graphs
show how close each case came. Signals below are hints for a listener, not findings.
Matching fingerprints
Each dot is a fingerprint hash found in both tracks. Shared audio forms a straight diagonal run
(highlighted); scattered dots are chance matches.
Offset agreement
How many matches agree on each time offset (reference time minus query time). Shared audio needs
one spike of at least 40 aligned hashes.Short samples
Aligned hashes along the query
Bars: hashes per second that agree with the best offset. Line: running total, reset after 2 s with nothing
aligned. A run must reach 40 to count, so a very short or sparse sample may never get there.Pitch-shifted samples
Pitch and speed sweep
Best aligned count after shifting the query's fingerprints by each number of semitones, pitch only or
pitch with speed (vinyl-style). Motif's detector only uses the 0 column; a tall bar elsewhere suggests a
pitch-shifted sample.Heavily processed samples
Share of query fingerprints matched
Per second: the share of the query's hashes found anywhere in the reference (bars), and the share on
the best offset (line). Filtering, distortion or time-stretching lower both.Dense mixes
Spectral peaks per second
Motif keeps at most 50 peaks per second. Above the line, quieter layers (such as a sample buried under
drums and vocals) are dropped before fingerprinting.
Melodic match by transposition
Best 8-second chroma similarity with the query transposed by each number of semitones. Musical
resemblance needs at least 0.8 for 3 or more consecutive windows.
Melodic match by tempo
The same similarity for each tempo ratio Motif tries. A cover played faster or slower peaks away from 1.0.
Pitch-class profile
How much of each pitch class each track uses overall, the basis for the key estimate. Covers in a
different key show the same shape shifted sideways.