Backend Analysis as the Foundation of Sound Generation
At the beginning of every process lies the continuous analysis of the audio material in the backend. This analysis operates independently of musical categories such as keys, scales, or harmonic theory. Instead, the incoming audio signal is evaluated in short time intervals—typically every 10 to 20 milliseconds. For each of these time frames, the algorithm produces two central values: an estimated fundamental frequency and a so-called confidence value.
The confidence value is of particular importance, as it expresses how certain the algorithm is about its own frequency estimation. Crucially, the algorithm never produces “no result.” Even in cases of silence, noise, or complex sonic events, a frequency estimate is always generated, accompanied by a correspondingly low confidence value. This design decision is conceptually central, as uncertainty is not treated as an error but as an integral component of the system. Uncertain analysis is not filtered out; instead, it is later used deliberately as sonic material.
fritsch music 