coherence-animation-networking

Sync Animator triggers by calling SetTrigger locally on the authority and broadcasting a coherence Command with MessageTarget.Other so remotes call SetTrigger too. Triggers are one-shot events, so late joiners correctly do not replay them.

Late joiners in T-pose means persistent animator state was never synced. Sync bools, floats, and ints as [Sync] properties that read and write Animator parameters directly, so joining clients receive the current values and their animator transitions naturally.

A CoherenceSync schema component holds at most 32 synced fields. For complex humanoids, split parameters into multiple sibling MonoBehaviours on the same prefab, grouped by concern such as locomotion, combat, and emotes, each getting its own 32-field budget.

Use a Command for one-shot events like a jump trigger, and [Sync] for persistent state like IsGrounded or IsDead. A [Sync] bool toggled for one frame is an anti-pattern; a trigger that late joiners must observe should be modeled as persistent synced state.

Coherence cannot cleanly sync IK positions or ragdoll results as fields. Instead, sync the inputs (look targets, death impulse via Command) and let each client compute IK or run the ragdoll simulation locally and deterministically.

Adding or removing [Sync] fields requires re-baking the coherence schema. Code that compiles will still misbehave at runtime against a stale schema, so always re-bake after changing synced field declarations.