Microvolts,
read inside the ear.

Neuraudio is an AI-designed EEG ear product. Consumer EEG has always been a dedicated device you put on for a session — a headband for sleep, a cap for the lab. We put the electrodes inside the earbuds you already wear: dry, continuous, decoded on the device.

Neuraudio picks up microvolt-scale EEG with dry in-ear electrodes, denoises and decodes it on the device, then closes the loop back into audio. No gel, no headband, no round trip to the cloud.

Synthetic source 500 SPS0.5 – 95 Hz50 Hz notch CYMBA C1 · 4 channels per ear
This waveform is generated live in your browser. Its components match the project's in-ear EEG simulator: 1/f pink noise + alpha bursts + mains hum + blink and jaw-EMG transients. It is not a recording and not human data.

System

Acquisition · Denoising · Closed loop

Acquisition

Acquisition

Silver-fabric dry electrodes sit on the cymba conchae, the cavum conchae and the mastoid, with the reference behind the ear. One ADS1299-4 per ear at gain 12 — not 24, because a dry electrode's DC offset commonly runs 50–200 mV and gain 24 would rail.

Denoising

Denoising

A 50/60 Hz notch plus a 0.5–95 Hz bandpass, all causal: online processing never touches future samples, or every chunk seam becomes a discontinuity. Artifacts are flagged by amplitude threshold and band-power gating rather than quietly deleted.

Closed loop

Closed loop

ANC means the speaker is never silent — the trouble with ear EEG, and also its cure. The noise-cancelling subsystem already carries feedforward and feedback microphones; those two plus the IMU are exactly the known references artifact cancellation needs. Adaptive cancellation against a known reference is mature DSP. Four-channel blind source separation is not.

CYMBA C1

First hardware

Concha-fit, with a stem. 25% taller than AirPods Pro, and the reason is stated plainly: the electrode blade has to span from the cymba to the cavum, and AirPods never have to do that. This is size that comes from function, not from weak design.

Channels8 (4 per ear)
Sample rate500 SPS
Passband0.5 – 95 Hz
Decoding band0.5 – 40 Hz
EMG monitor band55 – 95 Hz
Analog front endADS1299-4 × 2 · gain 12
Input-referred noise0.23 µVrms
Driver⌀10 mm dynamic
MicrophonesMEMS × 3 per ear
Active noise cancellationFeedforward + feedback
Inertial measurement6-axis per ear
End-to-end latency< 200 ms
Per-earbud size38.7 × 27.5 × 19.0 mm
Size options3 blades × 3 arms = 9 combinations

Teal marks neural-domain parameters, bronze marks acoustic-domain parameters.