Spatial Interaction Model — visionOS
Eye is intent. Hand is commit.
A spatial UI built by people who have not used a Vision Pro looks like a floating iPad — buttons reached for, hovers translated from web, persistent toolbars. A spatial UI built by people who understand the platform looks like peripheral attention plus deliberate confirmation.
Scroll through the six states of the canonical “select a transit route” interaction.
State 00
Resting
Trigger: Ambient state — no attention paid to any route.
All routes render at 60% opacity. Spatial audio is a low, slow city tempo bed at -36 LUFS.
Audio · ambient bed · -36 LUFS
State 01
Look-Target
Trigger: Eye gaze dwells on a route for ~120ms.
Route rises to 100% opacity over 280ms (Settle spring). Route ornament appears at the geometric centroid in Meridian Glass. Ambient bed drops 3dB; a positional tone emits from the route's anchor.
Audio · bed -3dB · positional tone -42 LUFS
State 02
Pinch-Commit
Trigger: Thumb-and-index pinch (hand can rest in lap; targeting is gaze).
Other routes drop to 22% over 240ms (Decisive spring). Ornament expands into a full information panel. Sonic confirm plays — 340ms, -28 LUFS, spatially anchored. Vehicle particles brighten ×1.4.
Audio · sonic confirm · 340ms · -28 LUFS
State 03
Pull-to-Detach
Trigger: Pinch + outward hand translation.
Ornament transitions from object-anchored to user-anchored. A faint Meridian Light thread connects window to route in space.
Audio · tether-thread resonance · -38 LUFS
State 04
Voice + Gaze Compound
Trigger: User speaks: "Show today versus last Tuesday."
Gazed-at route is the implicit subject. Comparison overlays without losing the spatial anchor.
Audio · voice confirm · faint sub-pulse
State 05
Dismiss
Trigger: Look away + pinch on empty space, or voice "Done."
Selection releases over 360ms. Ornament collapses, dissolves after -2s of no attention. Audio returns to State 0.
Audio · return to ambient bed · -36 LUFS
Hard rules
- No floating buttons in space. All controls are ornaments anchored to objects, or windowed panels.
- No iPad-translation gestures. Reach-and-tap is permitted only for direct manipulation under 0.5m.
- No imported hover treatments. Look-state is quieter, slower, and less saturated than hover.
- Spatial audio is not optional. Every state change has documented audio — position, intensity, duration.
- Spatial spring (mathematically critically damped) is mandatory. Vestibular safety, not stylistic choice.
- Accessibility substitutions (head-tracking, dwell, voice) are part of the deliverable, not afterthoughts.