Procedural, not generative: how MonRig animates a creature
MonRig is not a video model guessing motion. It computes every frame from your creature's anatomy, and that is why you can control it.
When people hear “automatic animation” in 2026, they assume a generative model somewhere is guessing the motion. MonRig doesn't work that way, and the difference matters when the creature is going into your game.
Every frame is computed
When you place a skeleton in MonRig, each bone gets a role: hip, knee, front paw, tail, wing tip, jaw. The clips are generated from those roles by an animation engine: a gait clock decides when each foot lifts, the feet plant where the ground is, the spine and tail follow through, an attack winds up, strikes and recovers. The same engine walks a centipede, a giraffe and a knight.
Nothing is sampled or hallucinated. The same rig with the same dials gives the same animation every time, down to the last frame.
That is why there are dials
Because the motion comes from rules, every part of it can be a control: stride length, lift, lean, how far the tail whips, how heavy a landing is, how many strikes a claw flurry has. You change the dial and the preview updates live. There is no prompt to rephrase and no seed to reroll.
And why the result fits your model
The engine only ever moves the skeleton you checked on your mesh. Feet stay on the ground, loops close exactly, and the file you download is a standard FBX and GLB any engine imports. Before a clip ships, the bake measures it: it checks for skin tearing, feet sinking into the floor and loops that don't close.
Where AI does help
One optional step does use a machine-learning model: Auto rig proposes where the guided-rig markers should go. It only proposes. You look at the markers, move what is off and decide. The rigging and every animation after that are computed, not generated.
If you want to see it on your own creature, start free: rig one by hand and preview every animation in the editor.