any loop, your key
Drop a sample on it, pick a key and a scale, and the loop moves into that key live, with the chords intact. Not a pitch shifter: two notes sounding at once are detected separately and each moves by its own interval. Drums and noise are copied through untouched.
Every real-time pitch tool on the market is 12-TET. Producers working in maqam, makam, dastgah or shruti have no way to retune recorded audio into their own system without going offline into an editor. Inkey treats those tunings as first-class targets rather than as an afterthought.
The other half is more prosaic. A sampled loop is almost never in the key of the track it is being flipped into, and transposing the whole thing drags the drums and the noise along with it. Inkey moves only what is tonal.
01 — Auto Fit
Snapping every note to the nearest degree sounds like enough. It is not. Measured on a real C♯ minor loop against an A minor target, pure snapping applies a mean correction of 52 cents and 49% of decisions have a rival degree within 30 cents — diatonic to the target and harmonically random. Auto Fit instead rotates a salience-weighted histogram against the target and commits one interval, which is what preserves every interval inside the loop.
02 — Harmonic grouping
Peaks are assigned to at most eight sources so that every partial of one note shares one shift ratio. A note has to own its own fundamental, or a major triad gets explained away as a single virtual fundamental two octaves down and the whole chord moves as one block. Notes may share partials — a fifth's second harmonic is the root's third — and shared evidence counts at a discount.
03 — Refusal
An interval is committed only with at least three distinct pitch classes, a winning rotation that clearly beats an average one, and four consecutive evaluations agreeing. A source must persist two hops before it moves: a note lasts many hops, a chance alignment of noise peaks lasts exactly one. A drum bed produces eight distinct pitch classes — more than any real loop — and what rejects it is that no rotation explains it better than any other.
Every figure here is printed by Inkey’s own test suite on rendered audio. Nothing in this table is an estimate.
The maqam, makam and raga tunings are derived in the test suite from published jins interval structures rather than compared against themselves, and cross-checked against published cents. Everything in the DSP and the editor is written from scratch over the DISTRHO Plugin Framework, which is ISC licensed with no revenue cap and no EULA.
Walkthrough
Being filmed. It will show Inkey on real material with the controls moving, because that is the only demonstration worth anything for this kind of plugin.
hear it where it lands
A physical model of a small speaker being driven too hard, and the protection DSP trying to save it.
make it survive the internet
A real transform codec you can starve, plus generation loss and packet loss no encoder does on purpose.