What a state machine actually is
An animation state machine is a graph. Each node is a state, such as Idle, Locomotion or Jump, and each arrow is a transition with a condition attached: "speed above 0.1", "is grounded equals false", "attack button pressed". The engine evaluates the graph every frame, decides which state the character is in, and cross-fades between clips when a transition fires.
In Unity this system is called Mecanim, and you build it in the Animator Controller window with parameters, blend trees and layers. In Unreal Engine you work in an Animation Blueprint, whose AnimGraph contains state machines, blend spaces and, increasingly, Control Rig nodes for procedural adjustments. Both tools look intimidating at first; both become friendly once you have built one complete character.
A typical locomotion graph
Most workshops build this same sequence, because it covers nearly every transition problem you will meet:
- Idle. A breathing loop with subtle weight shifts. Entry state for the graph.
- Locomotion blend. A 1D blend space from walk to run, driven by speed. Later, a 2D space adds strafing using direction.
- Jump start. Fires on input. Short, front-loaded, often with root motion turned off so gameplay owns the arc.
- Fall loop. Plays while airborne for any length of time, so long drops look correct.
- Land. Chooses a soft or heavy landing from fall velocity, then hands back to idle or locomotion.
Workshop drill: build the graph with grey-box animations first. Swap in polished clips only when every transition fires reliably. Systems before polish saves days of rework.
Blend times and why they matter
Every transition has a blend duration. Too short and the character pops between poses; too long and it feels floaty, with feet sliding during the cross-fade. A good starting point many leads suggest is somewhere between 0.1 and 0.2 seconds for locomotion changes, and near-instant blends for responsive actions like attacks. The right answer depends on the pose difference, so you test, adjust and test again.
Matching poses at transition points helps enormously. If your run cycle's left foot plants on frame 0 and your stop animation also begins on a left plant, the blend becomes nearly invisible. Unity's sync markers and Unreal's sync groups let you align cycles that have different lengths, which is essential when blending walk into run.
Layers, masks and additive animation
Characters often need to do two things at once: run while reloading, walk while looking at a target. Layers handle this. An upper-body layer with an avatar mask (Unity) or a layered blend per bone (Unreal) lets an arm animation override the torso while the legs keep running. Additive animations, such as a breathing offset or a hit flinch, add on top of whatever is playing. Learning when to override and when to add is one of the most valuable intermediate skills.
Beyond state machines: motion matching
Large graphs become hard to maintain. Simon Clavet's 2016 GDC talk on motion matching in Ubisoft's For Honor popularised an alternative: instead of hand-built transitions, the system searches a large database of motion-captured movement each frame for the pose that best matches the character's current state and desired trajectory. Unreal Engine 5 now ships motion matching through its Pose Search plugin. Advanced workshops introduce it, but most leads still teach state machines first, because they make you understand what a transition needs.
Procedural fixes on top
Even a perfect graph cannot predict uneven ground. Foot inverse kinematics plants feet on slopes and stairs; look-at constraints turn the head toward points of interest. Unreal's Control Rig makes these adjustments animator-friendly, and Unity offers the Animation Rigging package. Treat procedural layers as the final polish on a solid authored base, not a replacement for it.
How to practise this week
Download a free character from a source with a clear licence, such as the Mixamo library or the sample characters shipped with Unreal and Unity. Build the five-node graph above. Record a short screen capture of you running, stopping, jumping and landing, then watch it at quarter speed. Every pop you find is a lesson, and bringing that capture to a workshop critique is the fastest way to grow.
