DFRobot announced on July 29 that the seeMote Cube — a handheld spatial input device for Apple Vision Pro — is now available at US$299. It offers 6DoF motion tracking, six programmable physical buttons, and haptic feedback, and DFRobot is shipping open-source developer resources alongside it: reference accessory files, USDZ assets, and the source for its seeMote Viewer app.
Why a controller, on the headset that sold “no controllers”
Vision Pro’s founding pitch was that your eyes and hands are the interface — look at a thing, pinch to select. It’s a genuinely good interaction model for browsing, and a frustrating one for sustained precision work. Eye-and-pinch offers no tactile confirmation, gets tiring across a long session, and gives you nothing to grip when you’re manipulating an object in space and want to feel the difference between “moving it” and “letting go.”
That gap is why Apple’s move to publish specifications for third-party motion controllers on visionOS matters, and why hardware like this is showing up now. The seeMote Cube is aimed squarely at the professional cases where that friction bites hardest: spatial design review, digital twin and model inspection, and training simulation — DFRobot is explicit that it’s not positioning this for clinical or safety-critical use. In all of those, a person spends an hour rotating, scrubbing, and annotating a model, and the difference between a pinch gesture and a button that clicks under your thumb compounds every single time.
The haptics are the point, not a bonus feature
Six programmable buttons are conventional. What makes the device interesting for anyone working on interaction design is the combination of a tracked physical object and a haptic channel. Hand tracking can tell the system where your fingers are; it can’t tell your fingers anything back. Adding a vibration on selection restores the confirmation loop that touchscreens and physical controls provide for free — and gives developers a way to signal a boundary, a snap-to, or a failed action without another visual overlay in an already-busy field of view.
The open-source reference files and USDZ assets are a small but meaningful gesture too: they let a developer put an accurate virtual model of the controller in the scene, which is what makes a tracked object feel co-located with your real hand rather than approximately near it.
What isn’t answered yet
The announcement is thin on specifics that will decide how well this actually works: battery life, tracking range and method, latency, and connectivity are all unstated. $299 is a serious accessory price for a device aimed at developers rather than consumers, and its usefulness depends entirely on how many visionOS apps adopt third-party controller support now that the spec exists.
Still, it’s a notable marker. The first wave of spatial computing sold the absence of hardware as the feature; the second wave is quietly discovering that hands need something to hold.