From the source
- Google's Antigravity added a voice mode, powered by Gemini Live, that lets developers talk to the coding agent in the Antigravity App and CLI instead of typing every prompt.
- The company says the change answers a specific friction: as agents take on longer tasks, steering them means remembering a detail, answering a permission prompt or checking on progress, and each of those pulls a developer back to the keyboard.
- Voice mode continues the conversation it was started in, so the agent already has the context of earlier typed messages, and it hears the user's voice directly rather than a transcript, so a misheard word does not become a misunderstanding.
- The agent answers out loud in real time and can be interrupted by talking, and it still reads code, edits files and runs commands in the conversation.
- Destructive actions such as deleting files, running commands with elevated privileges or force-pushing to a Git remote cannot be approved by voice and require clicking Allow on screen, though they can still be denied by voice.
- By default the agent stays quiet and speaks only when it has an answer or needs input; saying "keep me posted" brings updates about every 30 seconds, and voice mode starts quiet again each time it is turned on.
- Users can pick from nine voices and set speaking speed from 0.75x to 2x, and in the CLI voice mode starts with /voice live or F6.






