PartialFKR
Strip a sound to the minimum partials that still carry its character. Hear it as raw audio, or hand the skeleton — MIDI/MPE, Csound, SDIF, JSON — to whatever instrument you actually want to hear.
There are many applications built around sinusoidal analysis — most optimised for high-fidelity reproduction of the input, or clean modification of it. PartialFKR is not one of those.
It's designed for the aggressive reduction of a signal to its core components, exported in a format useful for composition. The aesthetic reference points are works like Gérard Grisey's Partiels or James Tenney's Three Indigenous Songs — spectral analysis of a source becoming raw material for an entirely different instrumental reality. You analyse something, strip it down until it barely holds together, and hand the skeleton to an ensemble.
The workflow is a sort of f'd up convolution: the onsets, contours, and phonetic gestures of a source signal refracted through whatever timbre receives the exported data. Hence the name.
Fidelity is not the goal. Characteristic preservation through reduction is.
Windows builds are unsigned — SmartScreen will warn you. Use More info → Run anyway if you trust the source. macOS builds are signed and notarised; Linux builds are unsigned as is standard.
Built with the Loris library by Kelly Fitz and Lippold Haken. Type set in JetBrains Mono and Major Mono Display. Mascot animation is hand-built SVG, no frameworks.