Read what the
mirror won't.
A rolling lab of face, voice, and behavioural experiments — each one a small mirror that turns a photo, clip, or 30 seconds of attention into a single specific signal about you.
AI Stress
Scan.
The one everyone's sharing this week. A 7-layer facial signal model reads micro-tension across your brow, jaw, and skin tone — then maps it onto a thermal aura that's uniquely yours.
Try the scanTrending now.
A 7-layer facial signal read of micro-tension across brow, jaw, and skin.
Read your fundamental, your formants, and the singer you sound like at 11 a.m.
Sub-pixel skin colour changes that quietly count your heart, breath, and battery.
All experiments.
A quick read pulled from the face library — a single signal turned into a number you can share.
A quick read pulled from the face library — a single signal turned into a number you can share.
A quick read pulled from the face library — a single signal turned into a number you can share.
A quick read pulled from the face library — a single signal turned into a number you can share.
A quick read pulled from the face library — a single signal turned into a number you can share.
A quick read pulled from the face library — a single signal turned into a number you can share.
A quick read pulled from the face library — a single signal turned into a number you can share.
A quick read pulled from the face library — a single signal turned into a number you can share.
A quick read pulled from the face library — a single signal turned into a number you can share.
A quick read pulled from the face library — a single signal turned into a number you can share.
A quick read pulled from the face library — a single signal turned into a number you can share.
A quick read pulled from the character library — a single signal turned into a number you can share.
How an experiment becomes a reading.
Every experiment in the Playground passes through the same three doors before it lands here. None of them is the model.
- 01
Signal first, story second.
We start from a published peer-reviewed signal — facial action units, vagal tone, formant tilt — and audit it for stability across skin tone, age, gender, and mic quality before anything ships.
- 02
Anonymised by default.
Photos and audio are processed on-device for 11 of 23 experiments, and discarded after 24 hours for the rest. Results are tokenised; no face is ever attached to a result link.
- 03
Each result reviewed by a name.
An MD, PhD, or licensed therapist signs off on the language a reading uses. The credit appears on every share card — and so does the limitation.
Have an experiment worth running?
If you've got a signal nobody is reading yet — bring it. We publish three reader-submitted experiments a month.