Iris — eye-signal decoder

on-device · webcam BCI

Decode a choice from your eyes

Iris watches your eyes through the front camera while words flash, and learns to guess a binary or small-set choice you make in your head — which of two cards you focus on, or which word from a list you secretly picked.

Real science powers it: the gaze-cascade effect and pupil/recognition responses. It trains a small model on your eyes. Honest accuracy ≈ 60–85% after training — clearly above chance, but not reading your thoughts. Read the research →

🔒 On-device only. Video never leaves your phone. Only numeric features and a tiny model are stored locally. “reset” wipes everything.

Set up tracking

Hold the phone steady at a comfortable distance, eyes well lit, screen brightness constant.

Pick Use camera for real eye reading. Calibration is optional — you can jump straight in.

Choose an experiment

🎯 Focus Duel

Two cards. You focus on one in your mind. I guess which — from your gaze cascade.

🔮 Word Oracle experimental

Pick one word from a list, keep it secret. Words flash; I find your target from pupil & recognition.

🧠 Pupil Resonance SOTA

Options flicker at different rates. Fix your eyes on the centre and just think about one — your pupil entrains to its frequency and I read it. Works instantly; Personalize teaches it your pupil for higher accuracy.

⌨️ Mind Type eyes-only

Communicate with just your eyes: look to point, hold to select, blink to delete. Predictive words finish your thought in a couple of glances.

point with gaze · dwell to pick · blink = ⌫ · works best after calibration

🧬 Neuro Decoder micro-signals

Teach it two distinct mental states (think each on cue). It learns your fixational micro-jitter, drift & pupil micro-fluctuation signature for each — then reads which you're in, and proves it's real with a permutation test.