OpenPose editor with a real 3D body

How to make a ControlNet pose

1. Open the Studio and start from a pose in the library, a random pose (R), or the default figure.
2. Drag hands and feet to pose with IK, or rotate single joints. Set the hand shapes in the Hands panel.
3. Frame the shot: orbit the camera, choose a field of view, and set the export size to match your generation (e.g. 1024×1024).
4. In AI maps (ControlNet), choose OpenPose, turn on Include hands if your model supports them, and click Export map. All maps (.zip) exports the render plus OpenPose, keypoint JSON, depth, normal and edges in one go.
5. Load the image in your ControlNet unit (preprocessor: none, since the map is already processed) and generate.

What you can export

• OpenPose: the COCO-18 body skeleton in the standard ControlNet colours, optionally with hands. Keypoints come straight from the rig, not from pose detection, so nothing is missed or misread.
• OpenPose JSON: the keypoint file other pose editors and pipelines import.
• Depth: white near, black far. Best when the body shape and overlap matter, for example crossed arms or a figure seen from below.
• Normal: surface directions, for models that use normal maps to control lighting and form.
• Edges: clean outlines for Canny or Lineart ControlNets.

Why a 3D OpenPose editor

• Foreshortening is correct. An arm pointing at the camera gets shorter by exactly the right amount; on a 2D canvas you have to guess.
• Any camera angle. Low angles, high angles and back views are a camera move, not a re-drawing of every point.
• Hands that make sense. Pick a fist, point or peace sign and the 21 hand keypoints follow real finger joints.
• Several people. Add up to four figures; the export contains one skeleton per person.
• Depth from the same pose. Combine OpenPose with depth for difficult overlaps where a skeleton alone is ambiguous.