Morpheus

Most wearable sensing ends at a chart. Morpheus V1 asks what happens if the signal goes somewhere instead — if a body's own electrical activity moves an object in the room, immediately, while you are still making the expression that caused it.

The robot has a wide bar display carrying a pair of eyes and a live waveform that serves as its voice. It looks back at you. That is the point: the loop closes visibly, and nothing pretends to be more than a machine responding to a signal.

Every input is a face or a head. The gesture set was kept to four, because a control scheme you have to remember is a control scheme that fails in front of an audience.

A Muse S headband streams four channels of brain activity and six axes of motion over Bluetooth Low Energy to a single ESP32-S3 board. A gesture engine runs a fifty-times-a-second control loop across that stream. An onboard microphone feeds the waveform voice, and two continuous-rotation servos in a differential drive do the moving.

The headband's Bluetooth protocol is not published. Real-time gesture decoding meant taking the connection apart — watching the traffic, mapping the structure, and writing a client from scratch that could parse the sensor stream fast enough to hold a fifty-hertz loop. Roughly no part of that was in a datasheet.

One board runs the whole robot: a dual-core Xtensa LX7 at 240MHz with 8MB of memory and 16MB of storage.

The robot has its own manual — systems architecture, hardware schedule and gesture reference — written and set to the same standard as the object it describes. A prototype nobody can operate without you in the room is only half a prototype.