rig markers

Where the auto-rig's eight markers go, dragged over the mesh. Mixamo's auto-rigger asks for eight points on a front view (chin, wrists, elbows, knees, groin) and works the skeleton out from them. markerField3d is that step: the mesh's front silhouette with the markers on it, dragged where the joints are. A tail tip, which a front view cannot see, gets a side view beside it. Symmetry is on by default, so a wrist placed on one side places the other.

It is an ordinary scene-panel control, so it works on a page, in the scene and in a headset alike. The auto-rig page uses it: pick a character, drag a marker, and the character is fitted, bound and walking again when you let go.

Usage

import { markerField3d, fitSkeleton, meshQuery } from 'tosijs-3d'

const field = markerField3d({
  positions, indices,        // the mesh, as triangles (objTriangles reads an OBJ)
  value: markers,            // or leave it out: guessMarkers places them
  tail: { x: 0, y: 0.5, z: -0.9 },   // a tail tip, or leave it out
  handleCommit: (markers, tail) => refit(markers, tail),
})

handleChange fires as a marker moves; handleCommit once, when it is let go. Fitting and binding take a fraction of a second, so they belong on commit.

The guess

guessMarkers(silhouette) reads the front silhouette:

• groin: the crotch, the first filled point going up the centre line;
• wrists and elbows: the widest row of the upper body (the arms of a T-pose);
• knees: the middle of each leg, a little over halfway from the floor to the crotch;
• chin: the narrowest row between the arms and the top of the head (the neck).

It is a starting point to drag from, not a fit. A tail that hangs between the legs hides the crotch, and an A-pose has no widest row of arms; both fall back to proportions.

Coordinates

Markers are in the mesh's own units and frame, as fitSkeleton takes them: the character faces +Z with its left on +X, so in the front view its left is on your right. The side view has +Z (the front) on the right.