Wear the microphone.
Not the computer.
- dual MEMS capture target
- hard record / mute control
- LED + haptic capture state
- onboard buffer + secure sync
One device. Two ways to listen. A private voice AI Core with a tiny detachable Pebble that goes where your phone should not have to.
Signal splits voice capture from voice compute. Clip the Pebble where it hears naturally. Keep the Core in your pocket, on the desk, or docked nearby.
Current AI recorders usually choose between tiny capture hardware that needs a cloud, or larger hardware that is awkward to wear. Signal separates the constraints.
Audio is designed to land in encrypted storage under per-device keys, not a vendor’s default cloud bucket.
Whisper-class STT and Piper-class TTS are mature enough to form the local voice layer on compact ARM hardware.
Larger models or external services remain explicit and optional—not an invisible prerequisite.
The launch architecture is deliberately composed from proven primitives rather than requiring a breakthrough chip, battery or model.
Engineering targets for prototype selection and validation—not final certified specifications.
We’re opening a small Founder’s Edition reservation list for people and teams who genuinely need private voice capture.
A$10 refundable deposit is planned. No payment is taken on this preview.