Selected compositions and recordings exploring maqām, machine-generated accompaniment, and the boundary between algorithmic and human musicianship.
Arab takht recordings and recordings of automatically accompanied vocal improvisations are available at mawaweel.com/AudioRecording.
Al-Ghawanmeh, F., & Smaïli, K. Andalusian Fragrance (SMT-Based Musical Accompaniment). NIPS (NeurIPS) 2018 Creativity Art Gallery, Machine Learning for Creativity and Design Workshop, Montreal, Canada.
A machine learning-based automatic instrumental response to vocally improvised (mawwāl) narrative poetry, featuring the Mawaweel music accompaniment application. The improviser Muhannad Alkhateeb used elaborative expressiveness, maqam modulation, and register to convey passion, doubt, anger, sorrow, and hope; the computer stood as a peer in translating these creative vocal sentences. Recorded in studio across several takes, with light reverb, EQ, and compression applied — though the model is robust for live performance as well.
Un-tuned natural sounds sequenced automatically within a modal structure. The audio samples are vocal expressions — laughing, breathing, crying — of the composer's son Ryan at two months old. The modal structure follows the Mawaweel accompaniment system presented in the composer's master's thesis: an automatic instrumental accompaniment to a given vocal improvisation, with the vocal track muted and instrumental samples replaced by these un-tuned natural sounds.
A piece for nāy and piano in twelve-tone serialism.