Picnic: Playful Music-Making with Everyday Objects and Machine Learning

We present Picnic, an interactive musical installation exploring Machine Learning (ML) as design material for embodied music-making. The system transforms everyday objects in a picnic basket into a loop-based sampler, allowing users to create rhythms by striking cups, plates, and cutlery. A microphone captures sounds that are classified in real-time by an ML model, triggering percussion samples, melodic layers, and bird calls that loop sequentially. Embracing ML’s inherent uncertainty, we intentionally used an underfitted classification model to create a playful and ambiguous music-making experience.