Wetour Robotics Ltd. (NASDAQ: WETO) released a demonstration of its Conductor wristband, a surface electromyography device that decodes wrist muscle signals into real-time 3D hand poses and converts hand gestures into on-screen text commands without a keyboard or touchscreen.

The device uses an 8-channel sensor operating at 250 Hz. The company says it trained its model using Meta's publicly available emg2pose dataset, downsampled from Meta's original 16-channel, 2 kHz research configuration to match Conductor's consumer-grade hardware. The model was then fine-tuned through transfer learning on data collected directly from the 8-channel band.

Wetour stated the underlying sEMG-to-hand-pose capability is shared, open research rather than a proprietary development. The company said its focus is on making that capability practical and deployable on consumer hardware.

The architecture uses a streaming, state-space model called Mamba, chosen for what the company describes as linear-time, constant-memory inference designed to run fully on-device. In the demonstration, the model was evaluated on gestures it had not been trained on.

Conductor is part of Wetour's Orchestra platform, which the company describes as an open, cross-device system. Wetour also positions Conductor as a potential data-collection endpoint for enterprise robotics applications, capturing human intent and gesture data at the edge while keeping raw signals on-device.

The company's enterprise Early Access Program is open to qualified partners and includes hardware samples, SDK access, and co-development support.

Wetour is headquartered in Austin, Texas. The company noted that references to Meta Platforms do not imply any affiliation, endorsement, or partnership.