MaqamAI

2025 · code, sound, text

Before a model can be evaluated for maqam fidelity, the people evaluating it must learn to hear the relevant intervals and melodic behavior.

The learning sequence begins with curated references, interval recognition, melodic phrases, and guided comparison before introducing generated outputs. Early scope centers on a small set of maqamat rather than claiming comprehensive coverage. The project connects the Localization Gap’s audit findings to human listening practice and gives HAH-WAS reviewers a shared perceptual foundation. Its concept and tuning rules are documented; recordings, cultural review, lesson design, and a functional prototype remain required for public release.

Cinematic Rail Type: 8-card secondary proof rail
Initial scope: selected lessons in Rast, Bayati, and Hijaz
Method: reference listening, interval recognition, phrase comparison, and model evaluation
Status: product concept and tuning research documented; reviewed audio curriculum pending