Open-Source Hand Exoskeleton TAKTO ONE: Replacing Camera Tracking with Magnetic Encoders

A Digital Twin of a Hand, Without Cameras

A developer on Reddit, currently preparing for their master's thesis defense, showcased an open-source hand exoskeleton built from scratch — TAKTO ONE. Its core selling point is simple: 12 finger joints, each equipped with its own magnetic encoder, and the virtual hand in the headset is driven entirely by these sensors plus two IMUs, rather than relying on camera tracking.

What does this mean? Even if a finger is occluded, hidden under a table, or blocked by the other hand, the virtual hand can still accurately reproduce its pose. In contrast, pure vision-based solutions almost inevitably lose finger data in occluded scenarios. For VR/AR applications that require fine hand interaction, this is a shift in approach worth paying attention to.

Technical Breakdown: Why Magnetic Encoders Are Worth Watching

Sensing Architecture

TAKTO ONE uses a tendon-driven structure, with 12 finger joints each independently equipped with a magnetic encoder, plus two IMUs to capture the overall orientation of the hand. All data is fused to drive the virtual hand model in the headset.

This approach bypasses several common pain points of the computer vision route:

• Occlusion robustness: Fingers the camera can't see are still read by the encoders
• Controllable latency: Sensors read directly, no frame-by-frame inference needed
• Stable precision: Not affected by lighting or background interference

Of course, the trade-off is that users need to wear a physical exoskeleton, so convenience can't compare to pure gesture recognition.

Full-Stack Open Source

The project has released complete resources on GitHub: CAD files, PCB designs, firmware code, and assembly guides are all included. The developer also provides a browser-based virtual twin experience, allowing anyone to try out the hand tracking online without any hardware (takto.one).

This combination of "full-stack open source + online demo" significantly lowers the barrier for reproduction and secondary development. Anyone with access to 3D printing and PCB prototyping services has a chance to replicate one.

Magnetic encoder is a type of sensor that uses changes in a magnetic field to measure angle or displacement. It typically consists of a permanent magnet and a magnetic sensing element (such as a Hall sensor or magnetoresistive sensor). When the magnet rotates with the joint, the sensor detects changes in the magnetic field direction, thereby calculating a precise angle value. Compared to traditional potentiometers, magnetic encoders have no mechanical contact, resulting in longer lifespan and less wear; compared to optical encoders, they are less susceptible to dust, oil, and lighting, and can be made smaller. In scenarios like hand exoskeletons where space is tight and joints are dense, magnetic encoders can be independently installed at each finger joint, directly outputting absolute angles without zero-point calibration, making them ideal for real-time, low-latency pose tracking.

IMU (Inertial Measurement Unit) is a sensor module that integrates accelerometers and gyroscopes, sometimes including a magnetometer. It estimates an object's orientation and motion by measuring acceleration and angular velocity. In TAKTO ONE, two IMUs capture the overall orientation of the palm and back of the hand, while 12 magnetic encoders handle the fine angles of each finger joint. This division of labor is key: IMUs provide a global orientation reference, encoders provide local joint details, and after fusion, the virtual hand can be fully and accurately reconstructed in space. The downside of IMUs is integration drift over extended use, but in this project, they only handle the overall hand orientation and are fused with the encoders' absolute angle data, so drift effects are effectively suppressed.

Tendon-driven is a biomimetic mechanical structure that mimics how human fingers pull bones via tendons. In TAKTO ONE, the actuators (such as micro motors) are not directly mounted on each finger joint, but are centralized in the palm or back of the hand, remotely pulling each joint via thin cables or flexible tendons. The benefit is a significant reduction in weight and volume on the finger portion, making the exoskeleton lighter and more similar to a real hand's form. At the same time, tendon-driven design brings engineering challenges: cable pre-tension, friction, creep, and long-term reliability all require careful design and validation. This is also why the article mentions that "the long-term reliability of the tendon-driven structure needs more validation."

What the Open-Source Hardware Community Can Do With It

The application possibilities for hand exoskeletons go far beyond VR controllers. Combined with the project's open-source nature, several directions are worth exploring:

Teleoperation and remote manipulation. Using TAKTO ONE to drive a robotic hand or dexterous robotic hand in real time has direct value in hazardous environment operations, precision remote control, and other scenarios. Open-source firmware means you can adapt different communication protocols yourself.

Rehabilitation and motion assessment. The exoskeleton itself has joint angle measurement capabilities, which can in turn be used for motion capture and quantitative assessment in hand rehabilitation training.

VR/AR fine interaction. Scenarios requiring per-finger tracking, such as grasping, pinching, and typing, are exactly where pure vision solutions fall short, and sensor-based solutions are naturally suited.

Data collection gloves. For collecting high-quality human hand motion data for robot imitation learning, the precision of 12 independent joint readings far exceeds common resistive data gloves.

Trade-offs of the Hardware Route

TAKTO ONE is not without limitations. Wearable exoskeletons face engineering challenges in comfort, calibration time, and durability. The long-term reliability of the tendon-driven structure also requires more validation.

But its value lies in providing a reproducible open-source baseline. In the past, hand exoskeletons mostly remained in lab papers with hardware details undisclosed; TAKTO ONE lays out CAD, PCB, and firmware for all to see, enabling more people to iterate on this foundation.

For developers interested in AI programming, embedded development, and VR interaction, this project is worth bookmarking. Even without directly replicating the hardware, its sensor fusion approach and open-source organization are instructive.

What's Next

The developer is preparing for their thesis defense and has publicly solicited community feedback on Reddit — "What would you do with it?" This open attitude itself deserves recognition. For open-source hardware to form an ecosystem, what's needed is exactly this kind of move from paper to community.