Can your voice reveal stress, depression or Parkinson’s disease?

When we speak, our voice carries far more information than the words we say. Speaking rate, pauses, pitch variation, voice stability and other acoustic features can all be measured and analyzed.

Virtuosis focuses mainly on how a person speaks rather than on what they mean. Its technology analyzes pitch, rhythm, voice quality and other acoustic characteristics.

The reason is that speaking engages several systems at once: breathing, vocal folds, articulation, motor control and neurological functions. Some physiological or psychological changes can therefore subtly alter the way we speak. A review published in Digital Biomarkers notes that diseases affecting the heart, lungs, brain, muscles or vocal folds can alter the voice, and that artificial intelligence opens new avenues for screening and remote monitoring.

Part of the appeal of this approach is its simplicity. No new, complex medical sensor is needed: Virtuosis states that its system works from a recording of at least 30 seconds of natural speech, via a web app or an API.

This could pave the way for far more accessible prevention and monitoring tools. A person could, for example, record their voice regularly to see how it changes over time. That data could then complement other health information.

The work presented by Virtuosis spans several areas. In mental health, the company studies the links between the voice and stress, anxiety and depression. It also works on neurodegenerative diseases, including Parkinson’s and Alzheimer’s.

On its clinical validation page, Virtuosis reports, for depression, a sensitivity of 77% and a specificity of 83% in a cohort of 1,933 participants, using the PHQ-9 questionnaire as the reference. For Parkinson’s disease, the company reports a sensitivity of 98% and a specificity of 96% in 1,621 participants, using the UPDRS scale as the clinical reference.

How to read these figures: sensitivity is the share of affected people the tool picks up; specificity, the share of unaffected people it does not wrongly flag. They must, however, be read in context.

Virtuosis says as much itself, and independent research points the same way: a study published in the Annals of Family Medicine on a different voice tool, involving more than 14,000 adults, sees potential in these technologies to help clinicians screen for depression, while calling for their effectiveness to be measured in real-world clinical use.

This point is essential. Voice analysis is not meant to replace a psychiatrist, a neurologist or a physician.

Virtuosis explicitly states that its biomarkers are designed to support screening or monitoring, and that they complement professional judgment rather than replace it: a voice result is not, on its own, a diagnosis.

Their value lies instead in making certain changes easier to spot and, when those changes are significant, in pointing toward an appropriate medical assessment. The same caution applies to any risk score in predictive health: a signal invites a closer look, it does not settle the question.

At MyTwin, our ambition is to gradually bring together different dimensions of health within a single patient digital twin: biological, physiological, clinical, behavioral and digital data. In that vision, the voice could become one data source among many.

On its own, information drawn from the voice remains limited. But combined with sleep, physical activity, cardiovascular data, lab results or medical history, it could help us better understand how a person’s health evolves.

That is one of the principles of the digital twin: rather than looking at each indicator separately, trying to understand how several signals change together over time. It is also why Virtuosis is one of MyTwin’s partners: Claire’s story, about not knowing where to start with prevention, shows the place a mental health tool can take in a first health check-up.

One of the major shifts in preventive medicine could be the move from care based mainly on a few one-off measurements to care able to observe many signals regularly, what is known as real-world data.

Here the voice has an obvious advantage: we already use it every day. Phones, video calls, voice messages and computers already have the microphones needed to capture this kind of signal.

Provided these technologies are properly validated, used with people’s consent and built into a genuine care pathway, they could become a new tool for prevention and monitoring.

To learn more about Virtuosis AI’s technology and research, visit the Virtuosis AI website, and watch the full conversation in the MyTwin Inside episode on the voice.

In the end, the podcast asks a simple question: what if our voice became one of the many signals that help our digital twin better understand how our health evolves?