Generate high-quality AI speech using ElevenLabs' premium models.
More stable = less emotion, more consistent.
Higher = closer to original voice, potential artifacts.