The idea
Several phones record a performer from different viewpoints and reconstruct a skeleton plus textured body approximation for animation tools.
How it works
- Arrange about four phones around a moving performer.
- Combine synchronized viewpoints to infer an animated skeleton and textured human approximation.
- Export motion into Blender, Unity, Roblox, or Cubacadabra without suits or markers.
AI’s role: Estimate body keypoints and track identity across views; ordinary geometry and animation code build and export a textured approximation.
First demonstration
Film a simple dance move from four phones, solve a skeleton over time, and export the motion into a basic 3D scene.
What to solve next
How reliably can the system recover self-occluded limbs and moving clothing across different body shapes and camera placements?