🎚️ Multi-Signal Classification Engine
How Genre Classification Works
Genre classification is rarely determined by one piece of information.
The sorter therefore combines several signals, with stronger evidence taking priority over weaker supporting evidence.
Signal 1 — Text & Keyword Markers
The strongest classification clues often come from information already present in the project.
The sorter can examine:
Song titles
Original filenames
Preset names
Machine names
Explicit terms such as dubstep, trance, DnB, house, or synthwave can provide strong evidence for a genre assignment.
Signal 2 — Musical Pattern Features
The sorter can also extract selected features from the project's pattern and note data.
Examples include:
Rapid percussion patterns that can provide supporting evidence for genres such as Trap or Drum & Bass.
Pitch movement and slide behaviour that can provide supporting evidence for bass-oriented styles.
Sustained notes and pad-like patterns that can provide supporting evidence for styles such as Ambient or Trance.
These are deliberately narrow rules rather than attempts to perform complete musical analysis.
They provide supporting evidence alongside the stronger textual and structural signals.
BPM is used as another supporting signal.
Tempo ranges overlap between genres, so BPM alone is not treated as proof of a particular genre. Instead, it can reinforce other evidence already found in the project.
Signal 4 — Artist Associations
Where a recognizable artist name appears in the available metadata or filename, known artist-to-genre associations can provide additional contextual evidence.
These associations carry less weight than direct information from the project itself.