The human voice is more than a medium for communication—it’s a rich, multidimensional signal that reflects our physiological and psychological states. In a compelling article published in Nature (Outlook Feature), researchers explore how artificial intelligence (AI) is being trained to detect subtle voice biomarkers that may indicate conditions ranging from heart disease and diabetes to Alzheimer’s and depression.

Dr. Yael Bensoussan, director of the Voice Center at the University of South Florida, has long recognized the diagnostic potential of vocal cues. Now, through the NIH-funded “Voice as a Biomarker of Health” initiative (through the Bridge2AI Program) – she is helping to build a 10,000-person voice dataset to train AI models for clinical use. This project spans about 50 collaborators across 15 institutions and is pioneering ethical frameworks for voice data privacy.

These technologies hold promise to transform telemedicine, clinical trials, and mental health care. Voice could soon become our most powerful diagnostic instruments.