Implementing Look-At Logic in Motion Matching

Implementing Look-At Logic in Motion Matching

In this lesson, you will learn how to layer procedural look-at offsets over motion-matched animations to keep your character focused on a target without breaking the fluidity of their locomotion.

Why Motion Matching Needs Help with Focus

Motion Matching is exceptional at selecting the most natural-looking animation based on your character’s trajectory. However, because it relies on existing animation data, it can’t inherently "look" at a specific moving object if that movement wasn't captured in your database. If you simply rotate the head bone to a target, the sudden movement will clash with the high-fidelity locomotion, creating a "snapping" or jittery effect.

The Procedural Layering Approach

To solve this, we use a Layered Blend per Bone or a dedicated Look-At node within your AnimGraph, but we must process it after the Motion Matching node. The secret to maintaining the flow is to apply the rotation offset using Inertialization or a gentle alpha interpolation.

Instead of snapping the bone directly to the target, we calculate the FindLookAtRotation from the character’s head to the target. We then pass this rotation into a transform modification node. The key is to blend the weight of this rotation based on the character’s current state.

// In your AnimGraph logic, you would drive the "Alpha" of the Look-At 
// node based on the character's movement intensity.
float CalculateLookAtAlpha(float Velocity) {
    // Reduce focus when running, increase when standing still
    return FMath::Clamp(1.0f - (Velocity / MaxWalkSpeed), 0.0f, 1.0f);
}

Best Practices for Stability

1. Bone Constraints: Always define limits on how far the neck and head bones can rotate to prevent the mesh from deforming unnaturally.
2. Smooth Interp: Never set the rotation value instantly. Use FMath::RInterpTo to allow the head to track the target with a slight, realistic delay.
3. Hierarchy: Apply the look-at logic after your Inertialization node. This ensures that the look-at pose is treated as a final, additive layer rather than part of the pose selection process itself.

Key Takeaway: To maintain natural locomotion, treat look-at logic as an additive, alpha-blended procedural layer applied after the Motion Matching pose selection, ensuring head movement respects the underlying body momentum.