A camera captures the instructor, the server turns the frames into a pose stream, and an avatar in Roblox or Unity repeats the movement in real time.
{
"v": 1, "frame": 1284, "timestamp_ms": 1725790412034,
"coord": "world_normalized", "joint_order": "spec_web_v1",
"joints": [[0.50, 0.95, 0.00], [0.50, 0.72, 0.02], [0.50, 0.55, 0.01], …],
"confidence": [0.98, 0.97, 0.96, …],
"tracking": {"status": "ok", "lost": false, "quality": 0.98}
}
{
"v": 1, "frame": 1284, "timestamp_ms": 1725790412034,
"coord": "world_normalized", "joint_order": "spec_web_v1",
"joints": [[0.50, 0.95, 0.00], [0.50, 0.72, 0.02], [0.50, 0.55, 0.01], …],
"confidence": [0.98, 0.97, 0.96, …],
"tracking": {"status": "ok", "lost": false, "quality": 0.98}
}Only the stream of skeleton coordinates: the frames stay on the pose server.
An ordinary webcam or a phone pointed at the instructor; light and full height in frame matter more.
There is one stream; the scene client — on the Roblox or the Unity side — applies the pose to its rig.
Camera, pose server and scene client — from there the avatar repeats the movement.
Una cámara capta al instructor, el servidor convierte los fotogramas en un flujo de poses y un avatar en Roblox o Unity repite el movimiento en tiempo real.
{
"v": 1, "frame": 1284, "timestamp_ms": 1725790412034,
"coord": "world_normalized", "joint_order": "spec_web_v1",
"joints": [[0.50, 0.95, 0.00], [0.50, 0.72, 0.02], [0.50, 0.55, 0.01], …],
"confidence": [0.98, 0.97, 0.96, …],
"tracking": {"status": "ok", "lost": false, "quality": 0.98}
}
{
"v": 1, "frame": 1284, "timestamp_ms": 1725790412034,
"coord": "world_normalized", "joint_order": "spec_web_v1",
"joints": [[0.50, 0.95, 0.00], [0.50, 0.72, 0.02], [0.50, 0.55, 0.01], …],
"confidence": [0.98, 0.97, 0.96, …],
"tracking": {"status": "ok", "lost": false, "quality": 0.98}
}Solo el flujo de coordenadas del esqueleto: los fotogramas se quedan en el servidor de poses.
Una webcam normal o un teléfono apuntando al instructor; importan más la luz y el cuerpo entero en el encuadre.
El flujo es uno; el cliente de la escena, del lado de Roblox o de Unity, aplica la pose a su rig.
Cámara, servidor de poses y cliente de escena: a partir de ahí, el avatar repite el movimiento.
Камера снимает инструктора, сервер превращает кадры в поток поз, и аватар в Roblox или Unity повторяет движение в реальном времени.
{
"v": 1, "frame": 1284, "timestamp_ms": 1725790412034,
"coord": "world_normalized", "joint_order": "spec_web_v1",
"joints": [[0.50, 0.95, 0.00], [0.50, 0.72, 0.02], [0.50, 0.55, 0.01], …],
"confidence": [0.98, 0.97, 0.96, …],
"tracking": {"status": "ok", "lost": false, "quality": 0.98}
}
{
"v": 1, "frame": 1284, "timestamp_ms": 1725790412034,
"coord": "world_normalized", "joint_order": "spec_web_v1",
"joints": [[0.50, 0.95, 0.00], [0.50, 0.72, 0.02], [0.50, 0.55, 0.01], …],
"confidence": [0.98, 0.97, 0.96, …],
"tracking": {"status": "ok", "lost": false, "quality": 0.98}
}Только поток координат скелета: кадры остаются на стороне сервера поз.
Обычная веб-камера или телефон, направленный на инструктора: важнее свет и полный рост в кадре.
Поток один; клиент сцены — на стороне Roblox или Unity — применяет позу к своему риггу.
Камера, сервер поз и клиент сцены — дальше движение повторяет аватар.