I keep running into the same wall every time I try to set up additive blending for upper body overlays on top of locomotion. On paper it's simple: split at the spine with an avatar mask, run your aiming or reload on the upper half, let locomotion drive the lower half. In practice it keeps turning into a mess.
The thing that bites me every time is the reference pose for additive animations. In Unity, when you set a clip to Additive and bake against a reference frame, that delta gets applied on top of whatever the base layer is doing — which is fine until your base layer is a blend tree outputting a pose significantly different from your additive clip's reference. The arm correction that looked perfect over a standing idle completely falls apart when the character is sprinting and leaning forward. Shoulders creep up, spine twist accumulates in weird directions.
I've tried two approaches and neither is fully satisfying:
- Bake additive against the locomotion blend tree output — more accurate, but now the clip is tightly coupled to that specific base pose. Change locomotion and you're re-baking everything.
- Use a fixed bind pose as reference — more portable, but you accumulate visible errors as the base drifts, especially in spine rotation and shoulder height during fast movement.
In Unreal it's a bit cleaner. The Apply Additive nodes give you more explicit control over layering order, but the same fundamental problem exists. Your additive delta is always relative to something, and that something matters a lot.
Curious what approach people are actually shipping with. Are you living with the reference pose imprecision, building corrective blends on top, or is there a workflow I'm completely missing here?
