Skip to content

Character Animation

TiXL can load rigged and animated glTF models, play their animation clips, blend and retouch poses, and even rig your own procedural geometry to a skeleton. This page walks through the whole chain — from loading an animated character to driving particles with its motion.

Loading a rigged model

Add a LoadGltfScene and point its Path to a .glb or .gltf file that contains a skeleton (for a quick test, the free Khronos sample models Fox.glb or CesiumMan.glb work well). Besides the usual mesh and material outputs, a rigged file fills three more outputs:

  • SkinWeights — how strongly each part of the mesh follows each bone.
  • SkeletonPoints — the skeleton's rest pose as points. Connect it to DrawPoints to see the joints floating in place over the model.
  • ResultSetup — the scene setup, which now also carries the skeleton and all animation clips found in the file. Most of the operators below take this as input.

Playing an animation

Three operators form the basic playback chain:

  1. [SampleGltfAnimation] reads a clip from the setup and outputs a pose — one point per joint. Pick the clip with ClipIndex; by default the clip follows the playback time, and Loop wraps it at the clip's end.
  2. [PoseToSkinMatrices] turns that pose into the matrices that move each bone.
  3. [SkinMesh] deforms the mesh with those matrices and the skin weights.

Wire it up like this:

  • LoadGltfScene ResultSetup → [SampleGltfAnimation] Setup and [PoseToSkinMatrices] Setup
  • [SampleGltfAnimation] Pose → [PoseToSkinMatrices] Pose
  • LoadGltfScene Mesh and SkinWeights, plus [PoseToSkinMatrices] SkinMatrices → [SkinMesh]
  • [SkinMesh] Result → your usual mesh drawing, e.g. DrawMesh with the material from LoadGltfScene

Press play and the character animates. Because [SkinMesh] outputs a regular mesh, you can also feed it into any mesh effect — scatter points on the animated surface with PointsOnMesh, run it through deformers, or use it in collision setups.

Clip time is measured in seconds. When you connect something to OverrideTime, that value is used as seconds instead of the playback time — handy for scrubbing a clip with an LFO, an audio level, or a MIDI fader.

Blending and adjusting poses

A pose is just a list of points, so poses can be mixed before they reach [PoseToSkinMatrices]:

  • [BlendPoses] crossfades between two poses — for example two [SampleGltfAnimation] ops playing different clips of the same character. An optional weight mask limits the blend to some joints.
  • [AdditivePose] layers a clip on top of another one, relative to the rest pose — a breathing or sway clip on top of a walk, for instance.
  • [OverrideJoint] nudges a single joint from graph inputs: aim a head with a value, twist a spine with an LFO, wire a joint to audio.
  • [RetargetPose] transfers a pose from one character to another by matching joint names, so one animation can drive several differently proportioned rigs.

All of these output a pose again, so they chain freely.

Rigging your own geometry

Geometry that never had a skeleton can be bound to one:

  • [BindToSkeleton] generates skin weights for any mesh (or point buffer) by measuring the distance from each vertex to the skeleton's bones in rest pose. Radius controls how far a bone's influence reaches, FalloffPower how softly it fades. The result plugs into [SkinMesh] exactly like weights loaded from a file — so you can, for example, bind a completely different mesh to the Fox's skeleton and let it run with the Fox's animation.
  • [SkinPoints] is the point-buffer version of [SkinMesh]: bind a particle cloud or point set with [BindToSkeleton] and it deforms along with the rig.

Building a skeleton from points

Skeletons don't have to come from files. [SkeletonFromPoints] turns a CPU point list — for example a spline read back with PointsToCPU — into a skeleton: each point becomes a joint, parented to the previous one (separator points split chains). Together with two more operators this rigs procedural geometry to procedural motion:

  1. Take a snapshot of your spline with PointsToCPU (leave UpdateContinuously off, trigger it once) and feed it to [SkeletonFromPoints] — this frozen shape is the rest pose.
  2. Read the same spline back continuously with a second PointsToCPU (UpdateContinuously on) and feed it to [PoseFromPoints] together with the skeleton — this converts the moving spline into a pose.
  3. Bind your mesh with [BindToSkeleton], then [PoseToSkinMatrices] → [SkinMesh] as usual.

Now the mesh follows the animated spline like a snake following its path.

Troubleshooting

  • The model renders but doesn't move. Check that the pose actually reaches [PoseToSkinMatrices] — without a pose it outputs the rest pose by design. Also check the operator status indicators; the skinning operators report missing or mismatched inputs there instead of failing.
  • The mesh follows the wrong clip speed. Clip time is in seconds while the timeline runs in bars, so the playback BPM affects how fast timeline time maps to clip time. Connect OverrideTime for exact control.
  • Parts of the mesh stay behind when binding procedurally. Increase the Radius on [BindToSkeleton] — geometry outside every bone's envelope snaps to the nearest bone, which can look rigid.
  • The model appears at the wrong position or orientation. Skinned meshes are posed entirely by their skeleton; draw the [SkinMesh] result directly rather than through a scene-node transform, and use a Transform afterwards to place it.

See also