The heartbeat is simultaneously a biological event, a medical signal, and a cultural object. This research investigates what happens when it becomes computational raw material — captured via ECG, translated into audio, and processed through machine learning systems trained on everything except a heartbeat.
The work begins with a documented translation pipeline: raw cardiac data exported from an Apple Watch, passed through five computational stages, sonified using two methods (one literal, one compositional), and run through seven pretrained RAVE generative audio models and two Korg Monotron analogue synthesizers. Three structurally different kinds of processing logic: code designed by the researcher, patterns learned from training data, and circuit physics operated by a human hand in real time; each transforming the same signal according to their own criteria. What the pipeline makes visible is the layered set of decisions that stand between a body and its sonic representation. None of them are neutral.
Four physiological conditions have been tested so far: resting sinus rhythm, post-workout recovery, high heart rate after exertion, and a recording the device’s own algorithm classified as a Poor Recording. Comparing the outputs across conditions and models produced measurable findings about how different systems read the same signal, and what each one does when the signal gives it less to work with.
The research develops through three experiments: The Archive (a systematic comparative study across participants and devices), The Encounter (a public sonic installation), and The Threshold (a durational performance enacting the separation between a body and its computational double), and is being submitted as a PhD proposal in artistic research.