Speak silently.
Decoded by AI.
A wearable reads your facial muscles via sEMG electrodes and converts silent mouth movements to text. Bone conduction whispers the response back. No sound in, no sound out.
Signal path
How silent speech works
Silent path — Ag/AgCl electrodes capture sEMG from the masseter, submental, and orbicularis oris muscles. The ESP32-S3 runs DSP filtering and MFCC extraction, then streams features over BLE 5.0 to your phone for on-device inference.
Voice path — When you can speak normally, the phone mic handles it directly. Downloadable ASR models (Whisper, Parakeet, SeamlessM4T) run entirely on-device. No wearable needed.
Input modes
Adapts to your environment
Silent
Mouth words without sound. sEMG electrodes read muscle signals from your face and jaw.
Whisper
Phone mic picks up whispers while sEMG reinforces signal clarity. Best of both pipelines.
Translate
Silently mouth one language, hear another through bone conduction. Private real-time translation.
Voice
Standard speech via phone mic. On-device ASR, no wearable needed.
50-200
word silent vocabulary
<500ms
inference latency
1kHz
sEMG sample rate
20hr
battery target
Comparison
Not another voice app
vs. Typing
Hands-free, eyes-free. No keyboard, no screen. Just move your mouth.
vs. Voice assistants
Silent both ways. Bone conduction means only you hear the reply. No one around you knows.
vs. BCIs
No surgery, no bulky headset. Surface electrodes on skin — non-invasive, invisible, available now.
Waitlist
Be the first to speak silently.
Early access to the sEMG dev kit and mobile app.