MazzikaAI: Real-Time Arabic Maqam Accompaniment via Streaming Text-to-Music
MazzikaAI has been developed by researchers as a knowledge-driven system to offer real-time accompaniment for Arabic maqam music, which is noted for its microtonal intervals, modal frameworks, and intricate call-and-response techniques. This innovation fills a crucial void in generative music models, which mainly focus on Western equal-tempered systems, leaving Arabic maqam underrepresented. By utilizing natural language as a control mechanism, MazzikaAI integrates live MIDI, gestures, and inferred harmony into constantly refreshed text prompts. It operates an unmodified streaming generator, Google Lyria RealTime, without the need for model adjustments. The system incorporates expertise on six fundamental maqamat and relevant ornaments, ensuring subsecond key-to-audio latency. This research, available on arXiv (2608.10360), underscores the promise of text-to-music models with specific control interfaces tailored for distinct musical traditions.
Key facts
- MazzikaAI is a knowledge-based system for real-time Arabic maqam accompaniment.
- It uses natural language as the actuator of a real-time control loop.
- It compiles live MIDI, gesture, and inferred harmony into continuously updated text prompts.
- It steers an unmodified streaming generator, Google Lyria RealTime, without fine-tuning.
- The system embeds expert knowledge of six core maqamat, characteristic ornaments, and ensemble dynamics.
- It maintains real-time responsiveness with subsecond key-to-audio latency.
- The research is published on arXiv with identifier 2608.10360.
- The system addresses the gap in generative music models for Arabic maqam music.
Entities
Institutions
- Google Lyria RealTime
- arXiv