What voice-biomarker screening is
Voice-biomarker screening analyses how a person speaks — pitch, pace, pauses, energy, loudness and prosody — to surface early signals associated with psychological state. It does not transcribe or interpret the content of speech. A short spoken sample is enough: MindVoice uses a 45–90 second check-in.
The output is a signal, not a label. It indicates where attention or support may be worthwhile, and it is designed to be read alongside human judgement, existing wellbeing programmes and agreed referral pathways.
How MindVoice does it
- Consent-led check-in. A participant records a brief voice sample after clear consent, with privacy guardrails in place.
- Signal read. Acoustic features are mapped to validated models for the five current signals: distress, stress, burnout, fatigue and mental strain.
- Private support. The participant receives a private self-care plan by email and access to practical exercises such as box breathing, the physiological sigh and grounding.
- Anonymised cohort view. The organisation sees a strain map by team, site, function or year group — never individual health data — subject to minimum group-size thresholds.
What the evidence shows
MindVoice has completed a clinician-supervised, ethics-approved Phase 1 programme in India (CTRI/2025/09/094592), with 1,023 Phase 1 participants, a broader validation base of 1,665 participants and more than 40 hours of voice-led signal data. Development is guided by a Clinical Advisory Board, and Mindei Labs is a member of the NVIDIA Inception Program. Performance is reported at screening level and communicated per signal and deployment context.
Where it fits — and where it does not
Voice-biomarker screening works as an upstream layer: it can make emerging risk patterns visible earlier and route people toward EAP, counselling, occupational health or clinical review. It is not a diagnostic device, an emergency service, an employee-surveillance product or a performance-management tool, and it is not used for hiring, firing, promotion, compensation or discipline.
Common questions
- What is a voice biomarker?
- A voice biomarker is a measurable feature of how someone speaks — pitch, pace, pauses, loudness, energy, prosody and micro-timing — that correlates with a physiological or psychological state. Voice-biomarker screening reads these acoustic patterns, not the words or meaning of what is said.
- Are voice biomarkers a diagnosis?
- No. Voice-biomarker screening is a screening and signalling layer. It can indicate that a person or a cohort may benefit from support or further assessment; it does not diagnose a medical condition and is not an emergency service.
- How accurate is voice-biomarker screening?
- Accuracy depends on the signal, the population and how the screen is deployed. MindVoice reports screening-level performance and communicates it per signal and context. Its models are backed by a clinician-supervised, ethics-approved Phase 1 programme (CTRI/2025/09/094592) and a broader validation base of 1,665 participants.
- Does language or accent affect voice-biomarker screening?
- It can. Speaking style, language, accent and recording conditions all matter. MindVoice is built and validated for real-world diversity, and cohort-level reporting with minimum group-size thresholds reduces the impact of individual variation.
- Is voice-biomarker screening private?
- With MindVoice, the browser demo processes audio locally and never uploads it. In a pilot, capture is consent-led, individual results are private to the participant, and the organisation only ever sees anonymised cohort signals.