Your video never leaves this device. Hand tracking and sign recognition run in your browser. Only text (recognized signs, and the sentence to speak) is sent to Google Gemini and ElevenLabs. If you turn on reply captions, the other person's voice goes to your browser's speech service.
Camera is off.
Recognized signs
Hold a letter steady until the ring fills, or perform a sign you taught.
New here? Letters work much better after a 40-second quick calibration. And whenever a letter is wrong, tap it and pick the right one: the app learns your handshape. It stays on this device.
0 taught signs: none yet · teach a sign
Recognition settings
J and Z need motion and aren't in the letter model; teach them as motion signs. Keyboard: Space = word break, Backspace = delete, Enter = speak.