Voice settings

Settings → Voice controls both directions of speech: what listens to you, and what talks back.

Speech-to-text (dictation)

• Engine: browser (zero setup) · Parakeet (offline/local, auto-installed) · OpenAI (key required) · realtime (live session).
• For realtime, pick the powering service: OpenAI realtime or Grok. The engines guide compares them.

Text-to-speech (replies)

• Toggle spoken replies on/off, and independently choose whether agent output is read aloud.
• Voice options: free OS voices (offline), OpenAI neural voices, or Grok TTS — each with rate/pitch controls where the engine supports them.
• Remote voices fall back to the OS voice on any failure, so speech never silently dies.

Good to know

• Keys involved (OPENAI_API_KEY, XAI_API_KEY) live in your local .env — see API keys.
• Voice works everywhere text does: dictated commands hit the same command bar routing.