Building voice agents with new audio models in the API.

→ View original post on X — @openai