01 · Problem
What needed solving
Many teams have a phone image, short clip or archive video rather than a carefully controlled capture set. Waiting for ideal input can block early asset triage and spatial exploration.
02 · Approach
How we're building it
The prototype accepts common image and video media, samples frames within a bounded budget, produces a deterministic PLY scaffold and includes provenance so the relationship between input and output stays inspectable.
03 · Finding
What we learned
A modest, honest ingest layer can still create value for previsualisation, museum documentation and robotics data inspection, provided it is not described as complete 3D or learned Gaussian reconstruction.
04 · Next question
What we are testing next
What media-quality signals should determine whether an input proceeds to a stronger depth, camera-estimation or optimisation stage?
05 · Timeline
Where it is
- 2026Public image and video ingest prototype released
- 2026Bounded frame sampling and deterministic PLY plus manifest established
- NextEvaluate stronger depth and camera modules against representative phone captures
Related in Spatial Intelligence