Kaiju Voice and AI Brain
Kaiju Voice is the hands-free control layer for Kaiju Hub. Open AI Brain to start a spoken conversation, ask questions about the current app state, or carry out supported actions without navigating every screen yourself.
AI Brain shows the active conversation and controls. The Conversations area keeps prior sessions available for review, while AI Builder contains the deeper project knowledge and agent workflows the assistant can use.
Controls and models
Start and conversation
Start the hands-free loop from AI Brain. Use the configured hotkey to speak, continue the conversation, mute speech, or skip the current reply.
Response style
Concise keeps spoken replies short. Explain adds more context when you want the assistant to talk through its work.
AI model
Choose the provider, model, and thinking level used for the voice conversation. The separate SDK model controls app-side agent work.
Voice and timing
Choose an installed text-to-speech model and compatible voice, then set the pause before reply from 800 to 5,000 milliseconds.
What Kaiju Voice can control
You can ask what settings and actions are available before changing anything. Kaiju Voice reads the relevant built-in documentation only when a request needs it, keeping ordinary conversations focused while retaining exact command and option details.
Projects and workspace
Open projects and sidebars, manage files, run builds, sign or upload apps, check for updates, and inspect verified results.
Work management
Work with tasks, project history, GitHub, chat sessions, conversations, and AI Builder context.
Creative tools
Create and manage Block Library items, Marketing assets, Quantum jobs, generated media, and automations.
Compound workflows
For multi-step requests, Kaiju Voice checks prerequisites first, executes in order, and stops on the first failed step instead of claiming success.
Safety and conversation history
- Destructive, irreversible, publishing, or billing-related actions require explicit confirmation.
- Secrets and API keys are never read aloud or included in the assistant's settings summary.
- Action replies reflect the result returned by the app; a request is not reported as complete when a step fails.
- Use Conversations to revisit sessions and clean up or archive voice history when it is no longer needed.