Product
How Weevia S1 is being built to capture hand motion as 3D data.
How it works
Weevia S1 watches natural hand motion through 185° lenses. The hand-tracking model that turns it into 3D joint positions is in development — the goal is hand-motion data that robots can learn from.

Product images are previews — the final product design may change.
A compact stereo infrared sensor bar captures the hand from two viewpoints. Our own model for tracking the hand and its finger joints in 3D is in development; the resulting motion is meant to be recorded and exported as data — no gloves, no markers, no controllers.
- Wide-angle hand tracking (in development)
- 185° fisheye lenses (manufacturer spec) give the sensor a wide view. The tracking model in development is meant to follow a hand across that view.
- Jitter filtering (in development)
- Tracked joints tremble a little from frame to frame. A time filter that steadies their positions is in development.
- No gloves, no markers (in development)
- The tracking model in development works from infrared images of the bare hand, so the design needs nothing worn.
What makes Weevia S1 different
185° Lens
185° fisheye lenses (manufacturer spec) for a wide view in front of the sensor.
Plug In over USB
A standard UVC device — no driver to install.
Jitter filtering (in development)
A time filter that steadies joint positions, in development.
Python SDK
Opens the device, splits both eyes out of one frame, pins the exposure and drives the IR illuminator.
90 fps stereo capture
Global-shutter (sensor spec) infrared sensors capture hand motion at up to 90 frames per second.
Specifications
Where a value was not measured by us, its source is stated next to it. Anything not on this list is not published yet.
- Sensors
- Stereo monochrome infrared
- Stream
- 1600 × 800, both eyes in one frame
- Per eye
- 800 × 800
- Other modes (listed by the camera, not opened)
- 2560 × 800, 1280 × 400, 800 × 400
- Pixel formats
- MJPG at 90 fps (measured at 1600 × 800); YUY2 up to 45 fps (listed, not opened)
- Sustained frame rate
- 90.5 fps over 51 min at 1600 × 800 MJPG (one unit with an added heatsink)
- Baseline
- 64 mm (design value)
- Lens
- 185° fisheye (vendor spec), mechanically fixed focus
- Illuminator
- 850 nm, continuous (vendor statement, not measured)
- Host interface
- USB, standard UVC — no driver to install
- Device identity
- No USB serial number
Not published yet
These have not been measured on production hardware. We are not going to estimate them.
- Depth accuracy
- No error figures, because no calibrated depth measurement campaign has been run.
- Illuminator output and range
- No illuminance, working distance or coverage numbers.
- Power and thermals
- No draw or operating-temperature envelope.
The hardware today
Lens
185° fisheye lenses (manufacturer spec) give the sensor a wide view of the space in front of it.
IR illumination
A built-in 850 nm (manufacturer spec) infrared illuminator lights the scene in front of the sensor, and is switched and dimmed from software.
Capture Rate
Two global-shutter (sensor spec) infrared sensors deliver a 1600×800 stereo frame at up to 90 frames per second.
Where we’re headed
- Near term
- Our near-term target is hand-motion capture for robot learning — the sensor, its hand-tracking model, and the capture software.
- Long term
- Longer term, we are exploring other uses of the same hand tracking.
Join the waitlist
Weevia S1 is launching soon. Tell us what you need and we’ll reach out first — with early pricing and availability.
