Realtime voice & Grok

Per-clip dictation is press-mic-speak-release. The realtime engine instead holds a live session that listens continuously — closer to talking to a colleague than operating a recorder.

Setting it up

1. Settings → Voice → engine: Realtime.
2. Pick the service that powers it: OpenAI realtime (OPENAI_API_KEY) or Grok (XAI_API_KEY).
3. Start the mic from the command bar and just talk; recognized commands run as you speak them.

Grok, three ways

• Realtime conversation — the live voice session above.
• Text-to-speech — Grok’s TTS can be the voice Mastery HQ answers with (Settings → Voice).
• Hermes brain — Grok can power the orchestrator’s planning, alongside the CLI and VPS options.

Good to know

• Realtime sessions stream audio to the chosen provider while active — for a fully local setup, use Parakeet instead.
• Keys go in the local .env; nothing is proxied through third-party servers of ours.