Can blending emotion and speaking style make 3‑D facial animation more realistic?
Current speech‑driven facial animation usually treats emotion and speaking style separately, which can make the result feel less natural and expressive.
What they did
The authors introduced HiFTalker, a system that merges emotion and speaking style signals in a layered way so both can be controlled together.
What they found
They claim the combined control improves realism and expressiveness, though the abstract does not provide specific numbers.