Voice calls and Live mode

Voice needs an OpenAI API key in Settings, even if your advisors use Claude.

Two voice engines

• Realtime (default): speech-to-speech over a direct connection to OpenAI. Replies start in under a second and you can interrupt anytime. It waits until you sound finished before replying.
• Standard: your speech is transcribed, answered by your chosen model (Claude or ChatGPT), then spoken back. It's slower, but uses your preferred AI for the thinking.

Switch between them on the call screen (advisors) or in the assistant settings (Executive Assistant).

Calling an advisor

On My Board & C-Suite, click the call button on an advisor's card. You can also preview and change each advisor's voice there. The transcript is saved as a meeting, and the call is summarized into the Company Brain when you hang up.

Fixing problems

• "Microphone access was blocked": click the lock or site icon in your browser's address bar, allow the microphone, and reload.
• It cuts me off: pause less mid-sentence, or use headphones so the AI doesn't hear itself. Realtime waits for a natural end of thought.
• It hears itself or echoes: use headphones, or lower your speaker volume.
• "Realtime voice isn't available": check that your OpenAI key works (Settings → Test) and that your OpenAI organization is verified and has credit. Standard mode works as a fallback.
• Names are misheard: add them to your Business Profile or Company Brain. Twelve Seat sends your company's names and terms to the speech recognizer.