What's Behind the Image?

What needed solving

Single-image 3D systems can produce convincing previews while leaving users unsure which regions are measured and which are inferred.

How we're building it

A user can describe a hypothesis about what lies behind an image, inspect depth, occlusion, lighting and confidence cues, then rotate an evidence-backed reveal. Unknown and extrapolated regions remain explicit.

What we learned

Generative 3D becomes more useful when curiosity leads to the next evidence-gathering action. The interface can suggest which additional photograph would reduce uncertainty instead of silently completing the scene.

What we are testing next

How should a spatial interface communicate the boundary between observed geometry, a plausible prior and a generated hypothesis to non-specialist users?

Where it is

1. 2026
   Depth-aware reveal and uncertainty presentation prototyped
2. 2026
   Hidden-side completion kept separate from the current evidence-backed experience
3. Next
   Test evidence labels and next-view suggestions with creative and technical users