fritsch music

The Sample Architecture as Sonic Memory

A defining feature of EKO is the systematic creation of an extensive sample archive derived from the original recordings themselves. This archive is not a mere by-product, but a central sonic memory of the engine.

First, so-called sub-samples are extracted. These are short audio excerpts taken from particularly stable, sustained single-tone passages. The purpose of these sub-samples is to capture the sonic core of a recording. They are stored together with extensive metadata, including exact frequency, geolocation, and recording year. Typically, up to five such sub-samples are generated per original recording.

From these sub-samples, sustained samples are then produced. This process is based on granular synthesis and aims to largely dissolve the temporal structure of the source material. What remains is a spectral texture that can theoretically be played indefinitely. These sustained samples are particularly suitable for layered, hovering sound textures.

Another sample category consists of waveform samples. In this case, individual oscillation cycles are extracted from the sub-samples and stored as wavetables. These represent, in a sense, the “sonic nucleus” of a sound and later serve as raw material for various synthesizers.

Of particular conceptual significance are the insecurity samples. These are created from all moments within a recording in which the confidence value of the analysis is particularly low. These brief and often inconspicuous fragments—silence, background noise, recording artifacts—are collected over the entire duration of the recording and combined into a single sample. Insecurity samples carry little informational content about the recording itself, but they make the analysis process audible.