I always liked the original, Klatt-based synthetic voices that actually sounded like vocoders. I would think we could avoid the uncanny valley by creating synthetic voices that deliberately sound like synthetic voices.