Uthana Video-to-Motion 3.0 — AI Mocap With Hand Refinement and Foot IK

Uthana Video-to-Motion 3.0 converts ordinary single-person video into editable 3D skeletal motion — markerless motion capture without a suit, a stage or a calibration session.

What 3.0 Fixes

Uthana says animators kept telling them about two problems they were fixing by hand: fingers and foot sliding. Version 3.0 targets both directly:

  • A dedicated hand refinement step for finger articulation
  • Foot IK to lock planted feet to the ground

Those are the right two problems. Body motion from video has been usable for a while; hands and contact are where generated mocap has consistently fallen apart.

Why Hands and Feet Are the Hard Part

Hands are small in frame, frequently occluded by the body, and move fast. A pose estimator working from pixels has very little to go on, which is why AI mocap fingers tend to jitter or freeze in a generic half-curl.

Foot sliding comes from estimating each frame semi-independently. Tiny per-frame errors in the hip and leg chain add up, and a foot that should be planted drifts across the floor. It is the single most visible tell that motion was not hand-keyed or properly captured.

Foot IK addresses it at the root: detect contact, then constrain the foot to stay put while it is grounded.

Pricing and Output

  • $0.06 per second of output, pay as you go, credits do not expire
  • Export FBX, GLB or BVH
  • Retarget onto your own rigged biped, or auto-rig an unrigged mesh

Getting Good Results

Single person, full body in frame, steady camera, decent light, and clothing that does not hide the joints. Fast camera moves and partial occlusion remain the cases that break video mocap generally — shoot the reference with that in mind.

See our AI motion capture comparison for how it sits against the alternatives, including the offline option BlendCap.

Developer: Uthana · Price: $0.06/s pay as you go · Access: Web app, API, Blender and Maya plugins

Scroll to Top