A voice biomarker is a measurable feature of how you sound — pitch, loudness, pace, pauses, vocal clarity — that reflects a physiological or psychological state such as stress, fatigue, or energy. It captures how you say something, not the words you say.
How voice biomarkers work
When your nervous system shifts, your voice shifts with it — usually before you consciously notice. Stress tends to raise pitch and reduce vocal clarity; fatigue tends to slow your rate, flatten your tone, and add pauses. These changes are subtle and hard to hear yourself, but they are measurable. A voice biomarker system extracts those acoustic features from a short speech sample and maps them to a state.
Crucially, the strongest reads are relative: comparing today's voice against your own recent baseline ("higher or lower than your usual") is far more reliable than comparing you against the general population.
What HealthOS reads from your voice
HealthOS computes eight signals from a few seconds of unscripted speech: energy, stress, confidence, fatigue, vocal strain, expressiveness, articulation, and breathing. Each is a transparent, deterministic formula — a weighted blend of acoustic features compared against your roughly 30-day rolling baseline. It is not a black-box model guessing a number.
Is it grounded in real science?
Yes. The feature-to-state mappings come from decades of peer-reviewed speech research rather than being invented. HealthOS composes validated markers — for example, pitch and vocal-clarity changes under stress, and rate, tone, and pausing changes under fatigue. The honest limits matter too: phone-microphone noise means values are best read day-over-day within one person, not across different people, and the weights that combine markers are research-grounded priors still being calibrated against ground truth.
Why on-device matters
Most established voice-biomarker technology is clinical and cloud-based — your audio is uploaded for analysis. HealthOS runs entirely on-device: the features are extracted on your phone and your voice audio never leaves it. That privacy model is what makes a consumer voice-biomarker app possible in the first place.
Voice biomarkers vs. wearables
Wearables like Oura and WHOOP measure physiology (heart rate, HRV, sleep) and infer stress from it. Voice biomarkers read psychological and nervous-system state — confidence, vocal strain, expressiveness — that a wrist sensor cannot detect. The two are complementary. We go deeper on this in our voice vs. wearables comparison.
Evidence base draws on established speech-science research, including work by Scherer; Juslin & Laukka (vocal emotion); Krajewski (fatigue in speech); and Pennebaker (LIWC linguistic analysis).