Open a model at 0xr.space and press Zero, the microphone on the dock. Ask out loud: what is this model, who made it, find more like this. Zero answers by voice in the language you speak, and acts on the model: it searches your device and the Gallery, opens what it finds, points at named parts, isolates or hides them, pulls the assembly apart, cuts a section, scales it and plays its animation. In a headset it can also put a note, numbered steps or a comparison in the room.
Your speech goes to Google's Gemini Live service to be understood and answered, so this part does not run on the device. In mixed reality, if the camera is already allowed, Zero can see what you are looking at. Zero is not available in every country. It cannot make a real 3D model from a description: to explain an idea it shows a few simple shapes. It knows parts by the names in the file, and the screen from what the app reports.
Yes, as a preview with no sign-in and a daily allowance per device. Your own Gemini API key in Settings lifts it.
It replies in the language of your last full sentence and starts in English. A word or two does not switch it.
To Google's Gemini Live service, which turns it into the answer. Files you open are not sent; Zero is told the model's name and its part names. See the privacy page at https://0xr.space/privacy/ for what the camera features send and keep.
No. It can show a rough idea from simple shapes, but not build a 3D model to keep.