It learns your calm face, then watches six signals while you answer.
Demo needle — nothing is being read yet.
Twenty calm seconds to learn your face, then ask away.
For entertainment. Tellmeter measures visible stress, which is not proof of lying. Calm liars and nervous honest people both exist.
Tellmeter needs the front camera to read facial micro-signals. Video is processed on this device, frame by frame, and never leaves it. Nothing is uploaded, stored, or recorded.
Loading the on-device face model…
The face model downloads once, then runs offline forever.
Twenty calm seconds build a personal baseline from six facial and motion signals. Everything after that is measured as drift from that baseline — not against some average stranger.
A shrug, a scratch, a passing cloud — any single spike gets discounted hard. The index only climbs when several independent signals move together, and it's pulled back toward neutral whenever the camera read is poor.
The face model runs offline, inside the app. No frame, no landmark and no score is ever uploaded — there's no account and no server. Readings you save live in local storage and you can wipe them any time.