Skip to content

Voice

blu’s voice is a full-duplex realtime model (OpenAI GPT-Live). It listens, hands what you said to blu, and speaks the blobs’ answers and the outcome of every task.

  • ⌥⇧Space, the microphone button in the command bar, or the menu bar menu.
  • The menu bar icon shows when the microphone is listening.
  • The microphone is on only when you turn it on. Audio is never saved.

Echo cancellation lets you talk without headphones: the model doesn’t hear itself.

Settings ▸ Voice ▸ Talk while holding a key makes ⌥⇧Space a walkie-talkie: blu hears you only while the keys are held. Pressing it while the voice is off turns the voice on.

Talk the way you’d type: “Blue, make a menu”, “Bot, what’s up?”. The addressed blob answers with an “oop-oop” right away, the voice confirms, and when the task ends it tells you the outcome by itself — including permission requests and errors.

These are three different things:

What happens
You start talking over the voiceit stops at once and listens; agents keep working
“Be quiet”, “hush”the voice goes silent; agents keep working
“Stop the task”the agent’s task is cancelled

Without a key or network the voice shows the reason, and text commands keep working. A dropped connection is retried a few times on its own; tasks never depend on the voice.