A Quantized Native Runtime for On-Device Semantic Audio Generation | Matteo Spanio, Antonio Roda | ResearchPod