Talking with the AI¶
You can type or talk, from any device. Wherever you are, it's the same AI with the same memory of you.
Typing¶
Type in the box on the Bridge, in Comms, or in the AI box on any deck, and press Enter. Answers stream in as the AI writes them.
Talking¶
| How | |
|---|---|
| Hold to talk | Hold the mic button while you speak, let go when you're done. On a Starship, hold its talk key (Right Ctrl on Windows and Linux, Right Option on a Mac). |
| Wake words | Turn on the ear (it's called Listening) and say an AI's name: "Hey Celestra…", "OK Nova…". The ear stays on, on every device you're signed in on, until you turn it off yourself. |
| Hey Ship | Always gets your usual AI, whoever that is. |
| The name alone | A distinctive name works on its own: "Nova, what's the weather?" The Computer needs "Hey" or "OK" first, since "computer" is an everyday word. |
| Follow-ups | Right after the AI answers you can keep talking without its name for a few seconds. |
| Conversation mode | Say "let's talk" or press the speech bubbles: no name needed until "that's all" or 5 quiet minutes. If you turn and talk to someone else, the AI waits instead of chiming in. |
| Stop it | Just start talking over it (turn that off in Ready Room → Your AIs → Voice if its own voice from your speakers keeps stopping it). "Stop" works too. |
Room devices (Satellites and Shuttles) listen for the wake words all the time. On a shared device, the AI knows who's talking from their voice once they've taught it; a voice it doesn't know is a Guest.
Calling an AI that isn't your usual one, it checks first ("Nova here. Are you looking for me, or Celestra?") and your answer picks who. It doesn't ask again for a few minutes.
Sending a file¶
The paperclip (or dropping a file on the chat) sends a file with your message, up to 100 MB. Say what to do with it:
- "Read this": a PDF, Word document or text file goes on your reading shelf and the AI answers about it.
- "Keep this": any file goes to your Files in Ops, as it is.
- A picture: the AI looks at it and answers ("what does the sign say?"). This needs a Vision model, chosen in Models.
Send one on its own and the AI asks what you'd like, with a button for each thing it can do. Big videos, pictures and sound go to your Cargobay instead.
What the AI can do while you talk¶
Just ask, in your own words. A few examples:
| Ask | It |
|---|---|
| "What's the weather tomorrow?" | checks the forecast for where you live |
| "Look up the latest on the Mars mission" | searches the web and answers from what it found, with the links under its reply |
| "Read this page" with a link | reads it and answers about it |
| "Set a timer for 10 minutes", "wake me at 6:30" | sets it (a chip under its reply has Cancel) |
| "Remind me to call the dentist tomorrow at 9" | sets a reminder |
| "Add milk to my shopping list" | adds it |
| "Any emails from Kim today?" | checks your mail (Mail) |
| "Put the dentist on my calendar Tuesday at 3" | adds it (Calendar) |
| "Play a video about octopuses" | puts results on your Viewscreen |
| "Page Jordan in the kitchen: dinner's ready" | pages him there |
| "Away mission: compare three robot vacuums under $300" | works on it in the background (Away missions) |
| "What did we talk about Saturday?" | finds your past conversations (Comms) |
| "I want to talk to Lumina" | hands you over (The personas) |
What it does shows under its reply as a chip: a reminder you can cancel, a list, a command waiting for your Run, a tool's answer to copy, the links it used. Anything that changes something important asks for your OK, and sometimes your PIN, first.
When the AI says it did something, it did; when a tool doesn't answer, it says so plainly rather than making something up.