Bones are associated with characters, and that association keeps designers from using them where they would save the most time: interface elements that need to bend.
This is a practical guide to the cases, the workflow, and the limits.
What a bone rig actually buys you
A keyframe rig says: at 200ms, this vertex is here. At 400ms, it is there.
A bone rig says: this vertex is 60% influenced by the upper arm and 40% by the forearm. Move the arm and the vertex follows, everywhere, forever.
That indirection is the value. You animate the skeleton — a handful of rotations — and the geometry follows. Change the artwork, and the rig still works. Change the pose, and the skin deforms consistently.
For UI, that means:
- a mascot that leans, blinks and gestures without re-keyframing every path
- an illustration that squashes on press without distorting its neighbours
- a decorative arrow that bends along a curve instead of being redrawn
- a loading skeleton that breathes with a single parameter
- a hero object whose parts stay attached when the layout changes
The workflow, in the order it matters
1. Decide what the skeleton is, not what the drawing is.
Look at the artwork and ask: what are the levers? A tail has three. A ribbon has four. An arm has two plus a hand. Bones should be fewer than you think; a rig with twelve bones for a shape that needs three is harder to animate, not easier.
2. Parent the hierarchy properly.
Master bone at the root, children for the limbs or segments, each rotated around a sensible pivot. In Exode this is the master/child bone distinction in the hierarchy panel, and the pivot is what makes a rotation look like a joint rather than a spin.
3. Bind, then weight.
Auto-binding on selection is convenient and destructive if you are not careful — the audit notes on the Exode repo list exactly that problem and the fix that shipped. The safe sequence is: bind the mesh, then paint weights, then test extreme poses before you animate.
Weights are the craft. A vertex at 100% on one bone will crease. A vertex split 50/50 will bend smoothly. The batch weighting and the live weights table exist because doing this one vertex at a time is how rigs get abandoned.
4. Add IK only where you need it.
Inverse kinematics solves the pose from a target: drag the hand, the elbow follows. It is magic for a character reaching for something. It is overkill for a two-bone ribbon you will animate directly.
Exode supports IK chains; the rule is to use them at the end of a chain, not the root, or the whole rig becomes a puppet you cannot pose precisely.
5. Animate the skeleton, never the mesh.
Once the rig is bound, all animation should be bone rotations and root position. If you catch yourself keyframing individual vertices, the rig is wrong.
The UI-specific details nobody tells you
Layout and rigs interact badly if you let them. A bone rig lives in the coordinate space of its container. If the container reflows on a breakpoint, the rig must be positioned by layout, not by absolute coordinates, or it will detach on mobile. This is the single most common bug in "responsive mascot" work.
Rigs need a rest pose that matches the design. If the rig's neutral pose is not the design's default, every state has a permanent offset. Bind at the pose you will ship.
Reduced motion needs a rig-free fallback. A static image at the rest pose is the honest fallback. Do not run a cut-down rig animation for users who asked for less motion; show the still.
Export the rig, not a video. A baked sprite sequence loses the interactivity you built the rig for. If the rig is in the design artefact, the compiled component carries it.
Where the tooling is still catching up
The honest part.
Mesh deformation in Rive and Spine is deeper than in Exode today. If the brief is "a cape that ripples with wind", those tools have more mature free-form deformation and more artist-facing controls.
Exode's rigging audit lists eight gaps it closed — destructive auto-bind, single-vertex selection, read-only weights, edit-mode exit, static weight visuals, destructive re-binding, no independent Bézier handle weighting, and rigid normalisation. That is a real list, and the fact that it is public is a good sign. It is not the same as a decade of character rigs.
So the split I use:
- Character as art object, shipped inside a canvas or game: Rive or Spine.
- Character or shape inside a responsive interface, driven by app state: rig it where the layout and state live.
A small example
A brand mark that reacts to a form field:
-
idle: rest pose -
focus: slight lean toward the field, one bone rotation -
invalid: a single shake on the root bone, damped -
success: a squash-and-stretch envelope plus a small bounce
Four states, one rig, two bone rotations and an envelope. No keyframed paths, no per-state redraw, and the whole thing scales with the container because it is inside the layout.
That is the case for bones outside character animation: it turns "redraw it four times" into "rotate a lever".
The rule I would give a team
If an element needs to bend in more than one state, rig it. If it only ever translates or fades, keyframe it and move on. Bones are a multiplier on reuse, and reuse is the only reason they are worth the setup.
UX Baba writes about interaction design and the unglamorous parts of craft. Exode's bone system is part of the editor at exodeui.com.
