Librosa spectral detector (--detector librosa)¶

Uses spectral features (RMS energy, spectral centroid, zero-crossing rate) from librosa to classify segments.

Usage¶

praisonai-editor edit file.mp3 --preset songs_only --detector librosa

How it works¶

flowchart LR
    A[Audio frame] --> B[RMS energy dB]
    A --> C[Spectral centroid]
    A --> D[Zero-crossing rate]
    B --> E[Classify frame]
    C --> E
    D --> E
    E --> F{speech, music,\nsilence, singing?}

Strengths¶

• No extra install (librosa is a base dependency after install)
• Fast — runs on CPU without ML inference
• Works fine for clean recordings

Weaknesses¶

• Less accurate than ensemble or INA on noisy or complex audio
• May confuse speech over music as pure music