"Camera focus target switch causes jerk — smooth the target, not just the follower"

A chase camera uses exponential smoothing (Lerp(pos, desired, 1 - e^(-k*dt))) and follows the player smoothly during normal movement. But on a state transition — dismounting, exiting a vehicle, switching from one tracking mode to another — the camera jerks visibly, even though the position lerp is running the whole time.

The position lerp smooths the camera's output, but the input (focus target point and/or follow distance) jumps discontinuously on the state change. A chase camera that tracks an anchor at 1.5x distance while in one state and the body at 1.0x in another switches its desired target in a single tick. The position lerp then chases a target that jumped, and while it smooths the chase, the initial velocity kick reads as a jerk to the player.

Smoothing the follower doesn't help when the thing it's following teleports.

Give the focus point and follow distance their own exponential smoothing, separate from the position lerp:

// Per-state raw values
Vector3 rawFocus = isSwinging ? anchor.WorldPosition : body.WorldPosition;
float rawDist = isSwinging ? 1.5f : 1.0f;

// Smooth the target itself (slower rate than position lerp)
float targetRate = 5f;
_smoothFocus = Vector3.Lerp( _smoothFocus, rawFocus, 1f - MathF.Exp( -targetRate * Time.Delta ) );
_smoothDist = MathX.Lerp( _smoothDist, rawDist, 1f - MathF.Exp( -targetRate * Time.Delta ) );

// Position lerp chases the smoothed target (faster rate)
float posRate = 12f;
Vector3 desired = _smoothFocus - cameraDir * _smoothDist;
WorldPosition = Vector3.Lerp( WorldPosition, desired, 1f - MathF.Exp( -posRate * Time.Delta ) );

The target rate (~5) should be slower than the position rate (~12) so the focus point glides across the state switch and the position lerp never chases a discontinuity.

Seed _smoothFocus and _smoothDist to their raw values on the first valid frame, or they'll glide in from the origin on spawn.

Two nested exponential smoothers at different rates create a cascade: the slow outer smoother eliminates the discontinuity in the input, and the fast inner smoother (position lerp) tracks the now-continuous target without velocity spikes. The state change still happens — the camera still reframes from anchor to body — but it happens over a smooth arc instead of a single-tick snap followed by an asymptotic chase.