This late-breaking demo presents progress toward a DJ performance system that enables real-time manipulation of musical style. Drawing on Neural Style Transfer (NST) concepts, the envisioned system uses a hardware controller to interpolate between learned stylistic embed-dings during live performance. As an initial step, we adapt Magenta RealTime (Magenta RT) [1] — an open-weights, low-latency music generation model — for playlist-driven, culturally specific fine-tuning. Using a curated 88-minute playlist of Indian music, we explore a workflow where DJs upload a playlist as the dataset and fine-tune the model based on the style they intend to per-form. In informal listening tests, the fine-tuned model appeared to capture more of the melodic, rhythmic, and timbral qualities of Indian music compared to the origi-nal Magenta RT, suggesting the potential for personal-ized, culturally adaptive AI-driven DJ tools.