Set Speech to Speech output speed

Set audio.output.speed on session.update. Range is 0.7–1.5. Default is 1.0. Values below 1.0 slow the assistant; values above 1.0 speed it up. Change it mid-session when the caller needs a slower compliance line or a faster IVR prompt.

import json

await ws.send(json.dumps({
    "type": "session.update",
    "session": {
        "voice": "eve",
        "instructions": "You are a helpful assistant.",
        "audio": {
            "output": {
                "speed": 0.9
            }
        }
    }
}))

ws.send(JSON.stringify({
  type: "session.update",
  session: {
    voice: "eve",
    instructions: "You are a helpful assistant.",
    audio: {
      output: {
        speed: 1.2
      }
    }
  }
}));

Pair speed with the same session.update block you already use for voice, instructions, and output format. Input format and transport stay independent.

Pitfalls

• Stay inside 0.7–1.5. Values outside that range are invalid for audio.output.speed.
• Speed changes the spoken playback rate. Transcripts still show the words the model produced.
• Connect at wss://api.x.ai/v1/realtime?model=grok-voice-latest (or a pinned grok-voice-think-fast-* name) before you send session.update.
• Console API credits are separate from SuperGrok's weekly pool on grok.com.