fritsch music

From Raw Data to Musically Usable Pitch Events

The raw data delivered by the backend is initially unsuitable for directly controlling musical events. For this reason, a further algorithm exists within the composition engine that generates discrete pitch events from the continuous frequency stream. This process is multi-stage and follows a clearly defined logic.

First, a confidence gate is applied. Only when the confidence value reaches a high threshold—typically around 0.9—is a potential tone accepted as stable. As soon as such a stable state is detected, the gate is temporarily deactivated in order to track the continuation of the tone. The detected frequency is then converted into a MIDI note, deliberately using floating-point values. This preserves the microtonal precision of the analysis and frees the engine from the constraints of an equal-tempered tuning system.

In the next step, the system checks whether subsequent frequency values remain within a defined tolerance range. This range is approximately plus or minus 0.7 semitones. If this range is exceeded several times in succession, the system interprets this as the end of the stable tone, and a note-off event is generated. The confidence gate is then reactivated, and the process begins anew. In this way, monophonic pitch events are produced that are musically usable but not shaped by music-theoretical assumptions.