OpenAI

OpenAI API — Developer API

Schema-guaranteed output, transcription with speaker separation, and a realtime speech-to-speech session protocol.

Developer API Vendor page read September 1, 2026 developers.openai.com

What you can do with it

  • Constrain a response to a JSON Schema so a required key is never missing and an enum value is never invented.
  • Transcribe audio, with a separate model that returns speaker-attributed output and another that returns word-level timestamps.
  • Run a live voice session over WebRTC, WebSocket or SIP on gpt-realtime-2.1.

Playbooks it appears in

In the reference layer

Models it runs

Where this leads