Left: what you do. Right: what the room sees — only what you declared, after a lossy channel. Runs entirely in this tab.
Extras need a loaded avatar: the stick figure is the humanoid body and nothing else.
| Data | Tracker delivers | Avatar can show | Transmitted |
|---|
A part is transmitted only if the tracker delivers it and the avatar can show it
(spec §2.1). That set is declared once as caps.sends = .
Everything else stays in its rest pose on the right and costs no bytes. The right-hand
camera frames the declared region. The stick figure has a part for every data type:
fingers, one toe piece per foot, lids, mouth, tongue and pupils. What a loaded model
can show is read from the model. Finger joints are worked out on the receiving side
from the finger block (spec §5.5); the face comes from the expression block (§5.6).
Poses. Written as joint angles in the spec's conventions (§3.4) and grouped by
the part of the body they show. Pick a group, then one pose of it or a cycle through
the group; "All" cycles through every pose. The last group holds the conformance pose
vectors from testvectors/poses, legs only.
Turning. Drag on either view to turn both cameras around the avatar, or let "Turn" do it; a double-click puts them back. That is how a tail or wings get seen.
Extras. With a model loaded, the two lists hold its further bones and the expressions of its own that the face block does not carry (spec §2.6). Picking a name declares it; a chip takes it out again. Another avatar starts with nothing declared.
Hands, feet and face. "Count to ten" raises one hand and opens it finger by finger, thumb first, then the other hand. "Toes" lifts the toes of each foot, then stands on tiptoe. "Face" goes through the lids, the mouth shapes, the tongue and the gaze one at a time; the step that is on is named next to the selector. The view selector moves both cameras close to the hands, the feet or the face; on "auto" these three poses do that themselves. On a model, the toes are easiest to see barefoot, and the tongue direction has nothing to drive: only the stick figure shows it.
What moves. In every pose the hands and the face do what fits it: hands hang relaxed, lie flat on the lap, grip the hips or open for a wave; the mouth moves when the pose speaks or shouts. Otherwise only the lids blink and the eyes follow the head.
Hips height. The sender estimates how high the hips are from the lowest point of the
legs (spec Appendix D) and transmits it as h. On the high stool and in the air
nothing touches the floor, so those poses supply the height themselves.
Send rate is how many pose frames leave the sender each second (FPS). Spec tiers: 30 FPS (FULL) looks smooth; 15 FPS (NORMAL) still looks fine because the receiver interpolates between frames; 5 FPS (MINIMAL) is the readable floor. (The spec calls this rate in Hz; it's the same number.)