Rig and animate

rig_edit creates armatures, binds skins, repairs weights and authors actions. Every call needs a lease_id and an OBJECT-mode session.

Actions

Action	Use
`create`	New armature from 1–128 `bones` and an optional `location`
`bind`	Automatic weights from `name` armature to 1–16 `names` meshes
`weights`	Clean, limit and normalize weights on one skinned mesh
`pose`	Apply bounded pose transforms, optionally resetting first
`keyframe`	Create a new action from 1–32 keyframes
`clear_pose`	Reset every pose bone, optionally at a frame
`retarget`	Bake a source armature action onto the target rig
`set_action`	Assign an existing action, or clear the active action
`rename_action`	Rename an action
`delete_action`	Delete an unreferenced action
`nla_push`	Move an action into an NLA track/strip
`nla_remove`	Remove one strip, one track, or every track

Bones

Bone specifications take name, head, tail, parent, connect, deform and roll_degrees.

Armatures require unique names, nonzero lengths, acyclic parents, and connected bones that meet their parent tail. Bone names are limited to 63 UTF-8 bytes and armatures to 128 bones.

Binding and weights

Automatic binding requires 1–16 single-user meshes without modifiers or shape keys, at most 50,000 vertices and 64 bones. Binding snapshots mesh data and restores it on failure.

weights affects deform-bone groups only. It is bounded to 100,000 vertices, 512 groups and one million assignments. max_influences is 1–32, clean_threshold is 0–1 and normalize controls normalization.

Posing and actions

pose applies rotation_degrees, location and scale to named bones. Manual pose edits are not keyed, and an active action re-evaluates on frame changes.

keyframe writes a new action with action_name. Frames must be strictly increasing. Each action is limited to 32 frames, 128 poses per frame and 2,048 poses total. Failure removes the new action and restores the previous action and pose.

Clip management

inspect_rig lists every action with an object slot in actions, in addition to active and NLA-linked clips.

set_action assigns action_name to the rig; omit it to clear the active action. rename_action changes the action name and refuses collisions. delete_action removes only actions with no active, NLA or other users; clear references first.

nla_push places action_name, or the active action, into track_name and strip_name at start_frame. Strips in one track must not overlap. After a push the active action is cleared and the clip stays available for export. nla_remove removes a named strip, a whole track, or every track when track_name is omitted.

Animation QA

inspect_rig can sample a clip. Set action to a clip name, or omit it to use the active action, and pass checks:

root_motion reports root-bone displacement, horizontal and vertical travel,

maximum travel and whether the clip loops.

loop compares first and last bone matrices for translation, rotation and

scale.

deformation reports evaluated mesh bounds, maximum dimensions, dimension

ratio, centroid movement and non-finite vertices.

samples is 2–32 and defaults to 8. The call restores the previous action, action slot and frame. Deformation sampling is bounded to two million vertex samples. Foot sliding and engine import quality remain unchecked.

Retarget

retarget bakes a source armature action onto a target armature by world-space bone transfer. It takes the target name, source armature, source_action, new action_name and optional mapping and samples.

Without mapping, bones with matching names are used. With a mapping, each source bone maps to one target bone, target bones are used at most once, and every named bone must exist.

By default every whole source frame is keyed, up to 1,000 frames. samples selects 2–256 evenly spaced frames instead. Armatures are limited to 512 bones, mappings to 128 pairs and the result to 100,000 keyframes. The result reports mapped and unmapped bones.

Verify

Run inspect_rig after binding, repair, retargeting or animation. It reports hierarchy, all object actions, active/NLA clips, slots and weight quality without changing pose or action.