Anonymous: Speakers get letters (A, B, C). Nothing saved. Contacts: Match against saved voiceprints. Unknown speakers stay anonymous.
Smart Formatting
Auto-capitalize, punctuate, format numbers
Profanity Filter
Replace profanity with asterisks
Tap Calibrate to show a numbered ruler on the
glasses. Count the highest visible line number and the last readable column
number, then set them below.
LinesChars
Tune how readily voices match. Merge higher = fewer
duplicate speakers (one person split into A/B/C). Accept
= how close a voice must be to a saved contact to be named. Talk and watch the readout.
Accept0.45
Merge0.40
Speakers detected:
—
last match: —
recent: —
🗣️ Speakers in this session
▼
No speakers detected yet.
👤 Saved Contacts (0)
▼
No contacts saved yet.
🔒 Voiceprints are stored on this device only. They never leave your phone.
📋 Export saves a backup file. Import restores from a previous export (merges, no duplicates).
🔔 Name Alert
▼
Wake word detection
⚠️ Re-upload your .ppn file after page refresh (not stored locally).
Go to console.picovoice.ai → Create wake word → type your name → Download .ppn for Web (WASM)
Enter your AccessKey above, upload the .ppn file, and activate the toggle
Raw Transcript
Waiting for audio...
Click Start Listening to capture from your microphone.
On G2 glasses: double-tap to pause/resume, scroll to review history.
Requires the WebSocket server running on port 8080.