Voice·PD
Parkinson's Voice Analyser
wav2vec2 CV AUC 0.972 Sign in ✦ Motor test
AI · Speech biomarkers · Research preview

Detect Parkinson's
from a single breath.

Record three seconds of your voice saying "aaaah." A multilingual self-supervised model analyses acoustic biomarkers linked to early vocal degradation — the kind a doctor can't always hear.

0.995
Subject AUC
94.2%
CV Accuracy
61
Subjects
< 3s
Latency
1

Provide a voice sample

Say "aaaah" steadily for 3–5 seconds. Quiet room, mouth ~10 cm from the mic.

Or record live
🔊 Voice Spectrogram
Mel Spectrogram
2

Analysis result

Your result will appear here after analysis. We'll show the probability, a confidence gauge, and the acoustic features the model used.

Parkinson's Probability 0.0%
Healthy (0%) Threshold Parkinson's (100%)
Confidence
—
Model
—
View extracted acoustic features
FeatureValue
Research / educational prototype only. This is not a medical device. Voice-based screening has limitations; clinical decisions must be made by qualified physicians.