LOCALIZATION GAP

2024 · code, sound, systems

Generative music systems can imitate a regional style while erasing the tuning structures that make the music culturally specific.

The research compares generated tracks with documented modal and microtonal references, combining computational signal analysis with culturally informed listening. Repeated errors are organized into a failure taxonomy so that model behavior can be compared across prompts, outputs, and time. The corpus includes hundreds of generated tracks, but every public result must retain model version, test date, source, method, and interpretive limits. HAH-WAS and MaqamAI translate the audit into evaluation and education tools.

Cinematic Rail Type: 16-card paired rail
Corpus: more than 400 generated-track records under active audit
Method: structured prompts, reference listening, spectral comparison, and failure classification
Status: ongoing research; public results require timestamped methods and limitations

Related: hah was · tebr · maqamai · space time tuning machine